[Fix] (git) Added the ignored item .gradle

Initialize the desktop replay Request from the three iterationRP environment flags before loading MobileGL. Without this, the Request defaults caused the replay core to unset CI's exported flags, leaving all repairs disabled despite the workflow configuration.

.github/scripts/fetch-trace-fixture-lfs.sh

@@ -1,6 +1,10 @@
 #!/usr/bin/env bash
 set -euo pipefail
 
+script_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+# shellcheck source=trace-fixture-lib.sh
+. "${script_dir}/trace-fixture-lib.sh"
+
 if [ "$#" -lt 1 ] || [ "$#" -gt 2 ]; then
   echo "usage: $0 <trace-case> [fixture-dir]" >&2
   exit 2
@@ -62,57 +66,6 @@ if [ "${case_name}" = "OpenRA" ]; then
   exit 0
 fi
 
-get_lfs_metadata() {
-  local file="$1"
-  local pointer
-  local expected_oid
-  local expected_size
-
-  if ! pointer="$(git show "HEAD:${file}" 2>/dev/null)"; then
-    echo "failed to read tracked fixture metadata: ${file}" >&2
-    return 1
-  fi
-  if ! grep -q '^version https://git-lfs.github.com/spec/v1$' <<< "${pointer}"; then
-    echo "tracked fixture is not a Git LFS pointer: ${file}" >&2
-    return 1
-  fi
-
-  expected_oid="$(awk '$1 == "oid" && $2 ~ /^sha256:/ { sub(/^sha256:/, "", $2); print $2 }' <<< "${pointer}")"
-  expected_size="$(awk '$1 == "size" { print $2 }' <<< "${pointer}")"
-  if ! [[ "${expected_oid}" =~ ^[0-9a-f]{64}$ ]] || ! [[ "${expected_size}" =~ ^[0-9]+$ ]]; then
-    echo "invalid Git LFS pointer metadata: ${file}" >&2
-    return 1
-  fi
-
-  printf '%s %s\n' "${expected_oid}" "${expected_size}"
-}
-
-verify_fixture_file() {
-  local downloaded_file="$1"
-  local display_name="$2"
-  local expected_oid="$3"
-  local expected_size="$4"
-  local actual_oid
-  local actual_size
-
-  if [ ! -f "${downloaded_file}" ]; then
-    echo "fixture file is missing: ${display_name}" >&2
-    return 1
-  fi
-
-  actual_size="$(wc -c < "${downloaded_file}" | tr -d '[:space:]')"
-  if [ "${actual_size}" != "${expected_size}" ]; then
-    echo "fixture size mismatch for ${display_name}: expected ${expected_size}, got ${actual_size}" >&2
-    return 1
-  fi
-
-  actual_oid="$(sha256sum "${downloaded_file}" | awk '{ print $1 }')"
-  if [ "${actual_oid}" != "${expected_oid}" ]; then
-    echo "fixture SHA-256 mismatch for ${display_name}: expected ${expected_oid}, got ${actual_oid}" >&2
-    return 1
-  fi
-}
-
 fetch_file_from_mirror() {
   local file="$1"
   local url="$2"

.github/scripts/trace-fixture-cache.sh

@@ -0,0 +1,117 @@
+#!/usr/bin/env bash
+# Cache-side helper for trace fixtures.
+#
+#   key <case> [fixture-dir]     derive the actions/cache key and path list
+#   verify <case> [fixture-dir]  check restored fixtures against their pointers
+#   reset <case> [fixture-dir]   drop restored fixtures, leaving the pointers
+#
+# The cache key is content-addressed on the Git LFS pointer oids tracked at
+# HEAD, which are readable from a plain checkout without smudging. Fixture
+# content therefore maps 1:1 onto a key: unchanged content hits, changed
+# content is a new key and thus a miss, and the download path handles it. The
+# key deliberately carries no restore-keys prefix in the workflow - a fixture
+# that does not match the pointer exactly must never be restored.
+set -euo pipefail
+
+script_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+# shellcheck source=trace-fixture-lib.sh
+. "${script_dir}/trace-fixture-lib.sh"
+
+# Bump when the key derivation changes in a way that must invalidate old
+# entries; the content digest alone would not notice a format change.
+key_schema="v1"
+
+if [ "$#" -lt 2 ] || [ "$#" -gt 3 ]; then
+  echo "usage: $0 <key|verify|reset> <trace-case> [fixture-dir]" >&2
+  exit 2
+fi
+
+command_name="$1"
+case_name="$2"
+fixture_dir="${3:-tools/trace_replay/fixtures}"
+python_bin="${PYTHON:-python3}"
+
+if ! command -v "${python_bin}" >/dev/null 2>&1 && command -v python >/dev/null 2>&1; then
+  python_bin=python
+fi
+
+mapfile -t files < <(trace_fixture_files "${case_name}" "${fixture_dir}" "${python_bin}")
+if [ "${#files[@]}" -eq 0 ]; then
+  echo "no fixture files declared for trace case: ${case_name}" >&2
+  exit 1
+fi
+
+# Writes "name=value" to $GITHUB_OUTPUT when running under Actions, and to
+# stdout otherwise so the script stays runnable (and testable) off-CI.
+emit_output() {
+  local name="$1"
+  local value="$2"
+  if [ -n "${GITHUB_OUTPUT:-}" ]; then
+    if [[ "${value}" == *$'\n'* ]]; then
+      local delimiter="ghadelim_$(date +%s%N)_$$"
+      {
+        printf '%s<<%s\n' "${name}" "${delimiter}"
+        printf '%s\n' "${value}"
+        printf '%s\n' "${delimiter}"
+      } >> "${GITHUB_OUTPUT}"
+    else
+      printf '%s=%s\n' "${name}" "${value}" >> "${GITHUB_OUTPUT}"
+    fi
+  fi
+  printf '%s=%s\n' "${name}" "${value}"
+}
+
+sanitize_case() {
+  printf '%s' "$1" | sed 's/[^A-Za-z0-9._-]/_/g'
+}
+
+case "${command_name}" in
+  key)
+    manifest=""
+    for file in "${files[@]}"; do
+      # A case whose fixtures are committed directly rather than through Git LFS
+      # (OpenRA) has no pointer oid to key on, and nothing to download either.
+      # Report it as uncacheable so the workflow skips the cache entirely.
+      if ! metadata="$(get_lfs_metadata "${file}" 2>/dev/null)"; then
+        echo "trace case ${case_name} is not stored in Git LFS; skipping fixture cache" >&2
+        emit_output "cacheable" "false"
+        emit_output "key" ""
+        exit 0
+      fi
+      read -r expected_oid expected_size <<< "${metadata}"
+      manifest+="$(basename "${file}") ${expected_oid} ${expected_size}"$'\n'
+    done
+
+    digest="$(printf '%s' "${manifest}" | sha256sum | awk '{ print substr($1, 1, 16) }')"
+    safe_case="$(sanitize_case "${case_name}")"
+
+    emit_output "cacheable" "true"
+    emit_output "key" "trace-fixture-${key_schema}-${safe_case}-${digest}"
+    emit_output "paths" "$(printf '%s\n' "${files[@]}")"
+    ;;
+
+  verify)
+    for file in "${files[@]}"; do
+      metadata="$(get_lfs_metadata "${file}")"
+      read -r expected_oid expected_size <<< "${metadata}"
+      verify_fixture_file "${file}" "${file}" "${expected_oid}" "${expected_size}"
+    done
+    echo "Verified ${#files[@]} fixture file(s) for ${case_name} against the tracked Git LFS pointers."
+    ;;
+
+  reset)
+    # Put the working tree back to the pointer files a fresh checkout would
+    # have, so that a rejected cache entry falls through to exactly the same
+    # download path a cache miss takes.
+    for file in "${files[@]}"; do
+      rm -f "${file}" "${file}.tmp"
+    done
+    git checkout -- "${files[@]}"
+    echo "Reset ${#files[@]} fixture file(s) for ${case_name} to their tracked Git LFS pointers."
+    ;;
+
+  *)
+    echo "unknown command: ${command_name}" >&2
+    exit 2
+    ;;
+esac

.github/scripts/trace-fixture-lib.sh

@@ -0,0 +1,73 @@
+#!/usr/bin/env bash
+# Shared helpers for trace-fixture handling: reading the in-tree Git LFS pointer
+# metadata and verifying a fixture file against it. Sourced by
+# fetch-trace-fixture-lfs.sh (verify after download) and by
+# trace-fixture-cache.sh (cache key derivation and verify after cache restore),
+# so both paths agree on what a valid fixture is.
+
+# Reads the Git LFS pointer tracked at HEAD for a fixture path and prints
+# "<oid> <size>". Fails if the tracked blob is not a well-formed LFS pointer.
+get_lfs_metadata() {
+  local file="$1"
+  local pointer
+  local expected_oid
+  local expected_size
+
+  if ! pointer="$(git show "HEAD:${file}" 2>/dev/null)"; then
+    echo "failed to read tracked fixture metadata: ${file}" >&2
+    return 1
+  fi
+  if ! grep -q '^version https://git-lfs.github.com/spec/v1$' <<< "${pointer}"; then
+    echo "tracked fixture is not a Git LFS pointer: ${file}" >&2
+    return 1
+  fi
+
+  expected_oid="$(awk '$1 == "oid" && $2 ~ /^sha256:/ { sub(/^sha256:/, "", $2); print $2 }' <<< "${pointer}")"
+  expected_size="$(awk '$1 == "size" { print $2 }' <<< "${pointer}")"
+  if ! [[ "${expected_oid}" =~ ^[0-9a-f]{64}$ ]] || ! [[ "${expected_size}" =~ ^[0-9]+$ ]]; then
+    echo "invalid Git LFS pointer metadata: ${file}" >&2
+    return 1
+  fi
+
+  printf '%s %s\n' "${expected_oid}" "${expected_size}"
+}
+
+# Checks an on-disk fixture against the size and SHA-256 from its LFS pointer.
+verify_fixture_file() {
+  local downloaded_file="$1"
+  local display_name="$2"
+  local expected_oid="$3"
+  local expected_size="$4"
+  local actual_oid
+  local actual_size
+
+  if [ ! -f "${downloaded_file}" ]; then
+    echo "fixture file is missing: ${display_name}" >&2
+    return 1
+  fi
+
+  actual_size="$(wc -c < "${downloaded_file}" | tr -d '[:space:]')"
+  if [ "${actual_size}" != "${expected_size}" ]; then
+    echo "fixture size mismatch for ${display_name}: expected ${expected_size}, got ${actual_size}" >&2
+    return 1
+  fi
+
+  actual_oid="$(sha256sum "${downloaded_file}" | awk '{ print $1 }')"
+  if [ "${actual_oid}" != "${expected_oid}" ]; then
+    echo "fixture SHA-256 mismatch for ${display_name}: expected ${expected_oid}, got ${actual_oid}" >&2
+    return 1
+  fi
+}
+
+# Prints the fixture file paths of a trace case, one per line. Strips CR so the
+# result is usable when python emits CRLF (Git Bash on Windows).
+trace_fixture_files() {
+  local case_name="$1"
+  local fixture_dir="$2"
+  local python_bin="${3:-python3}"
+
+  "${python_bin}" tools/trace_replay/trace_cases.py \
+    --format fixture-files \
+    --case "${case_name}" \
+    --fixture-root "${fixture_dir}" | tr -d '\r'
+}

.github/workflows/apk.yml

@@ -11,6 +11,9 @@ on:
 jobs:
   build:
     runs-on: ubuntu-latest
+    permissions:
+      actions: write
+      contents: read
     env:
       CCACHE_BASEDIR: ${{ github.workspace }}
       CCACHE_COMPRESS: "true"
@@ -41,12 +44,11 @@ jobs:
           gradle-version: 8.10.2
 
       - name: Restore ccache
-        uses: actions/cache@v5
+        uses: actions/cache/restore@v5
         with:
           path: .ccache
-          key: ${{ runner.os }}-apk-${{ github.job }}-ccache-${{ github.ref_name }}-${{ github.run_id }}
+          key: ${{ runner.os }}-apk-${{ github.job }}-ccache-v1
           restore-keys: |
-            ${{ runner.os }}-apk-${{ github.job }}-ccache-${{ github.ref_name }}-
             ${{ runner.os }}-apk-${{ github.job }}-ccache-
 
       - name: Install ccache
@@ -125,6 +127,28 @@ jobs:
         if: always()
         run: ccache --show-stats
 
+      # Rewrite one rolling entry per job on the default branch. The upload stays
+      # cumulative - it carries every object restored at the top of this run plus
+      # the few TUs that actually changed - but Actions cache keys are immutable,
+      # so the superseded blob has to be released before the same key can be
+      # re-uploaded. Running after the build means a failed build leaves the
+      # existing entry untouched. The other trigger branches restore this entry
+      # rather than each writing a ~4 GB one of their own.
+      - name: Release superseded ccache entry
+        if: github.ref_name == github.event.repository.default_branch
+        env:
+          GH_TOKEN: ${{ github.token }}
+          CACHE_KEY: ${{ runner.os }}-apk-${{ github.job }}-ccache-v1
+        run: gh cache delete "${CACHE_KEY}" || true
+
+      - name: Save ccache
+        if: github.ref_name == github.event.repository.default_branch
+        continue-on-error: true
+        uses: actions/cache/save@v5
+        with:
+          path: .ccache
+          key: ${{ runner.os }}-apk-${{ github.job }}-ccache-v1
+
       - name: Verify APK metadata and packaging
         run: |
           AAPT2="$(find "$ANDROID_HOME/build-tools" -name aapt2 -type f | sort -V | tail -n 1)"
@@ -185,7 +209,7 @@ jobs:
       - name: Load trace cases
         id: trace-cases
         run: |
-          echo "android=$(python3 tools/trace_replay/trace_cases.py --ci --format github-apk)" >> "$GITHUB_OUTPUT"
+          echo "android=$(python3 tools/trace_replay/trace_cases.py --ci --format github-apk-matrix)" >> "$GITHUB_OUTPUT"
           echo "names=$(python3 tools/trace_replay/trace_cases.py --ci --format names)" >> "$GITHUB_OUTPUT"
 
   trace-fixtures:
@@ -201,9 +225,41 @@ jobs:
       - name: Checkout repo
         uses: actions/checkout@v6
 
+      - name: Derive trace fixture cache key
+        id: fixture-key
+        run: bash .github/scripts/trace-fixture-cache.sh key '${{ matrix.case }}'
+
+      - name: Restore trace fixture cache
+        id: fixture-cache
+        if: steps.fixture-key.outputs.cacheable == 'true'
+        uses: actions/cache/restore@v5
+        with:
+          path: ${{ steps.fixture-key.outputs.paths }}
+          key: ${{ steps.fixture-key.outputs.key }}
+
+      - name: Verify restored trace fixture
+        id: fixture-verify
+        if: steps.fixture-cache.outputs.cache-hit == 'true'
+        run: |
+          if bash .github/scripts/trace-fixture-cache.sh verify '${{ matrix.case }}'; then
+            echo "ok=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "ok=false" >> "$GITHUB_OUTPUT"
+            echo "::warning::Cached fixture for ${{ matrix.case }} failed verification; falling back to the download path"
+            bash .github/scripts/trace-fixture-cache.sh reset '${{ matrix.case }}'
+          fi
+
       - name: Fetch trace fixture
+        if: steps.fixture-verify.outputs.ok != 'true'
         run: bash .github/scripts/fetch-trace-fixture-lfs.sh '${{ matrix.case }}'
 
+      - name: Save trace fixture cache
+        if: steps.fixture-key.outputs.cacheable == 'true' && steps.fixture-cache.outputs.cache-hit != 'true'
+        uses: actions/cache/save@v5
+        with:
+          path: ${{ steps.fixture-key.outputs.paths }}
+          key: ${{ steps.fixture-key.outputs.key }}
+
       - name: Stage trace fixture
         run: |
           safe_case="$(printf '%s' '${{ matrix.case }}' | sed 's/[^A-Za-z0-9._-]/_/g')"
@@ -281,13 +337,7 @@ jobs:
     strategy:
       fail-fast: false
       max-parallel: 4
-      matrix:
+      matrix: ${{ fromJSON(needs.trace-cases.outputs.android) }}
-        backend:
-          - name: DirectGLES
-            gpu: software
-          - name: DirectVulkan
-            gpu: lavapipe
-        case: ${{ fromJSON(needs.trace-cases.outputs.android) }}
     steps:
       - name: Set Swap Space
         uses: pierotofy/set-swap-space@v1.0
@@ -370,6 +420,9 @@ jobs:
           MOBILEGL_USE_ANGLE: ${{ matrix.backend.name == 'DirectGLES' && '1' || '0' }}
           MOBILEGL_TRACE_ANGLE_VARIANT: ${{ matrix.case.name == 'minecraft-1.21.4-fabric-iris-bliss-in-world' && '90a62123d794' || 'ec889e6ea831' }}
           MOBILEGL_MAGMA_R11G11B10F_FALLBACK: ${{ matrix.backend.name == 'DirectVulkan' && '1' || '0' }}
+          MOBILEGL_FIX_ITERATIONRP_SUBGROUP_SCRATCH: ${{ matrix.backend.name == 'DirectVulkan' && matrix.case.name == 'minecraft-1.21.4-fabric-iris-iterationrp-in-world' && '1' || '0' }}
+          MOBILEGL_DERIVE_NUM_SUBGROUPS: ${{ matrix.backend.name == 'DirectVulkan' && matrix.case.name == 'minecraft-1.21.4-fabric-iris-iterationrp-in-world' && '1' || '0' }}
+          MOBILEGL_ITERATIONRP_FIX_BARRIER: ${{ matrix.backend.name == 'DirectVulkan' && matrix.case.name == 'minecraft-1.21.4-fabric-iris-iterationrp-in-world' && '1' || '0' }}
         run: |
           apk_file="android-retrace-apks/MobileGL-plugin-trace-release-${GITHUB_SHA}.apk"
           test -f "${apk_file}"
@@ -379,6 +432,9 @@ jobs:
           if [ "${{ matrix.backend.name }}" = "DirectGLES" ] && [ "${{ matrix.case.name }}" = "minecraft-1.21.4-fabric-iris-bliss-in-world" ]; then
             extra_retrace_args+=(--avoid-angle-llvmpipe-sampler-mipmap-min-filter)
           fi
+          if [ "${{ matrix.backend.name }}" = "DirectGLES" ] && [ "${{ matrix.case.avoid_angle_llvmpipe_explicit_lod_bias || false }}" = "true" ]; then
+            extra_retrace_args+=(--avoid-angle-llvmpipe-explicit-lod-bias)
+          fi
           if [ "${{ matrix.case.coherent_as_flush || false }}" = "true" ]; then
             extra_retrace_args+=(--coherent-as-flush)
           fi
@@ -411,6 +467,24 @@ jobs:
           run_retrace || retrace_status=$?
           if [ "${retrace_status}" -eq 75 ]; then
             echo "::warning::Android emulator infrastructure failed; restarting it and retrying this retrace once."
+            # Surface-lost is retried rather than failed, so it would otherwise
+            # be invisible. Report it per job - a healthy run prints nothing and
+            # a rate spike shows up as a row per affected case.
+            reason_file="android-retrace-result/infrastructure-failure-reason.txt"
+            surface_lost_retries=0
+            if [ -f "${reason_file}" ]; then
+              surface_lost_retries="$(grep -c 'angle-surface-lost' "${reason_file}" || true)"
+            fi
+            if [ "${surface_lost_retries}" -gt 0 ]; then
+              echo "surface-lost retries: ${surface_lost_retries} (${{ matrix.backend.name }}, ${{ matrix.case.name }})" \
+                >> "${GITHUB_STEP_SUMMARY}"
+            fi
+            # The restart truncates EMULATOR_LOG, and the attempt that lost the
+            # emulator is the one worth reading - the retry usually only shows
+            # the wreckage. Keep the first attempt's log before it is clobbered.
+            if [ -f "${EMULATOR_LOG}" ]; then
+              cp "${EMULATOR_LOG}" "${EMULATOR_LOG}.first-attempt" || true
+            fi
             sh android-plugin/run-avd-ci.sh stop \
               --avd-name "${AVD_NAME}" \
               --emulator-log "${EMULATOR_LOG}" \
@@ -450,6 +524,13 @@ jobs:
           if [ -f "${EMULATOR_LOG}" ]; then
             cp "${EMULATOR_LOG}" android-retrace-result/diagnostics/emulator.log
           fi
+          if [ -f "${EMULATOR_LOG}.first-attempt" ]; then
+            cp "${EMULATOR_LOG}.first-attempt" android-retrace-result/diagnostics/emulator-first-attempt.log
+          fi
+          # A vanished emulator looks identical whether the host OOM killer took
+          # qemu or the renderer faulted. These two say which.
+          free -h > android-retrace-result/diagnostics/host-memory.txt 2>&1 || true
+          sudo dmesg -T 2>/dev/null | tail -300 > android-retrace-result/diagnostics/host-dmesg.txt || true
 
       - name: Stop Emulator
         if: always()
@@ -531,23 +612,42 @@ jobs:
           )
 
           if ((${#failed_cases[@]})); then
-            echo "Retaining fixtures for failed retrace case(s):"
+            echo "Retaining fixtures and results for failed retrace case(s):"
             printf '  %s\n' "${!failed_cases[@]}"
           else
-            echo "All retrace jobs succeeded; no fixtures need to be retained."
+            echo "All retrace jobs succeeded; nothing needs to be retained."
           fi
 
           deleted=0
           retained=0
           while IFS=$'\t' read -r artifact_id artifact_name; do
+            keep=0
             if [[ "${artifact_name}" == MobileGL-trace-fixture-* ]]; then
               case_name="${artifact_name#MobileGL-trace-fixture-}"
               if [[ -v "failed_cases[${case_name}]" ]]; then
+                keep=1
+              fi
+            elif [[ "${artifact_name}" == MobileGL-android-retrace-result-* ]]; then
+              # The result artifact carries mobilegl.log, retrace.log, logcat,
+              # the emulator log and the actual/diff images - the only record of
+              # why a retrace failed. Its name ends in -<backend>-<case>, so a
+              # suffix match on the case name keeps both backends' results for a
+              # case that failed on either of them, which is what a comparison
+              # needs. The match is anchored at the end, so a case name that is a
+              # prefix of a longer one does not retain the longer one's results.
+              for case_name in "${!failed_cases[@]}"; do
+                if [[ "${artifact_name}" == *-"${case_name}" ]]; then
+                  keep=1
+                  break
+                fi
+              done
+            fi
+
+            if ((keep)); then
               echo "Retaining ${artifact_name} (${artifact_id}) for failed retrace."
               ((retained += 1))
               continue
             fi
-            fi
 
             echo "Deleting ${artifact_name} (${artifact_id})"
             gh api --method DELETE "repos/${GITHUB_REPOSITORY}/actions/artifacts/${artifact_id}"

.github/workflows/test.yml

@@ -1,4 +1,4 @@
-name: Test
+name: Test
 
 on:
   push:
@@ -11,6 +11,9 @@ on:
 jobs:
   build-linux:
     runs-on: ubuntu-latest
+    permissions:
+      actions: write
+      contents: read
     env:
       BUILD_DIR: build-linux
       CCACHE_BASEDIR: ${{ github.workspace }}
@@ -34,12 +37,11 @@ jobs:
         uses: lukka/get-cmake@v4.3.3
 
       - name: Restore ccache
-        uses: actions/cache@v5
+        uses: actions/cache/restore@v5
         with:
           path: .ccache
-          key: ${{ runner.os }}-test-${{ github.job }}-ccache-${{ github.ref_name }}-${{ github.run_id }}
+          key: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
           restore-keys: |
-            ${{ runner.os }}-test-${{ github.job }}-ccache-${{ github.ref_name }}-
             ${{ runner.os }}-test-${{ github.job }}-ccache-
 
       - name: Prepare Vulkan SDK
@@ -83,6 +85,8 @@ jobs:
             -DMOBILEGL_LOG_ACTIVE_LEVEL=MOBILEGL_LOG_LEVEL_INFO \
             -DMOBILEGL_BUILD_TEST=ON \
             -DMOBILEGL_BUILD_BENCHMARK=ON \
+            -DMOBILEGL_BUILD_INTEGRATION_TEST=ON \
+            -DMOBILEGL_ITEST_VK_ICD=/usr/share/vulkan/icd.d/lvp_icd.json \
             -DMOBILEGL_BUILD_TRACE_REPLAY=OFF \
             -DBENCHMARK_DOWNLOAD_DEPENDENCIES=ON \
             -DBENCHMARK_ENABLE_TESTING=OFF \
@@ -95,6 +99,28 @@ jobs:
         if: always()
         run: ccache --show-stats
 
+      # Rewrite one rolling entry per job on the default branch. The upload stays
+      # cumulative - it carries every object restored at the top of this run plus
+      # the few TUs that actually changed - but Actions cache keys are immutable,
+      # so the superseded blob has to be released before the same key can be
+      # re-uploaded. Running after the build means a failed build leaves the
+      # existing entry untouched. The other trigger branches restore this entry
+      # rather than each writing one of their own.
+      - name: Release superseded ccache entry
+        if: github.ref_name == github.event.repository.default_branch
+        env:
+          GH_TOKEN: ${{ github.token }}
+          CACHE_KEY: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
+        run: gh cache delete "${CACHE_KEY}" || true
+
+      - name: Save ccache
+        if: github.ref_name == github.event.repository.default_branch
+        continue-on-error: true
+        uses: actions/cache/save@v5
+        with:
+          path: .ccache
+          key: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
+
       - name: Package Linux runtime
         run: |
           mkdir -p ci-artifacts
@@ -110,6 +136,7 @@ jobs:
             "${BUILD_DIR}/CTestTestfile.cmake" \
             "${BUILD_DIR}/MobileGL/MG_Test" \
             "${BUILD_DIR}/MobileGL/MG_Benchmark" \
+            "${BUILD_DIR}/MobileGL/MG_IntegrationTest" \
             "${SHARED_LIBS[@]}"
 
       - name: Upload Linux runtime
@@ -159,12 +186,105 @@ jobs:
       - name: Test
         working-directory: build-linux
         run: |
+          ulimit -c unlimited
+          sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
           if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" = "true" ]; then
             ctest -V -L unit --no-tests=error
           else
             ctest --output-on-failure -L unit --no-tests=error
           fi
 
+      - name: Upload core dumps
+        if: failure()
+        uses: actions/upload-artifact@v7
+        with:
+          name: unit-core-dumps
+          path: /tmp/core.*
+          if-no-files-found: ignore
+
+  integration:
+    runs-on: ubuntu-latest
+    needs: build-linux
+
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Get CMake
+        uses: lukka/get-cmake@v4.3.3
+
+      - name: Install runtime dependencies
+        # Same set as the benchmark job, for the same reason: the scenarios bring
+        # up real headless EGL (llvmpipe) and Vulkan (lavapipe) contexts, and
+        # libegl-mesa0 - the EGL vendor library behind glvnd's libegl1 dispatch -
+        # only arrives as a Recommends.
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libvulkan1 libegl1 libegl-mesa0 libgles2 libgl1-mesa-dri mesa-vulkan-drivers
+
+      - name: Download Linux runtime
+        uses: actions/download-artifact@v8
+        with:
+          name: mobilegl-linux-runtime
+          path: .
+
+      - name: Unpack Linux runtime
+        run: tar -xzf mobilegl-linux-runtime.tgz
+
+      - name: Normalize CTest command paths
+        run: |
+          python - <<'PY'
+          from pathlib import Path
+          import re
+
+          for path in Path('build-linux').rglob('CTestTestfile.cmake'):
+              text = path.read_text()
+              text = re.sub(r'"[^"]*/cmake-[^"]*/bin/cmake"', '"cmake"', text)
+              path.write_text(text)
+          PY
+
+      - name: Integration scenarios
+        working-directory: build-linux
+        # REQUIRE_GPU makes a driverless runner FAIL instead of skipping every
+        # scenario - an all-skip run is otherwise indistinguishable from a pass,
+        # which is how a five-month-old draw-dropping bug survived unseen until
+        # this lane existed.
+        #
+        # The lavapipe ICD pin lives in the build-linux configure
+        # (-DMOBILEGL_ITEST_VK_ICD), NOT here: the configure bakes it into each
+        # test's ctest ENVIRONMENT property, and a property entry OVERRIDES the
+        # job environment - a VK_ICD_FILENAMES exported here would be silently
+        # ignored while looking like it works. This lane runs on lavapipe
+        # deterministically, not on whichever of the eight Mesa ICDs a GPU-less
+        # runner enumerates first.
+        #
+        # Cores are armed so that any crash - the harness pre-flight child's
+        # included - leaves /tmp/core.*, which the failure-only step below ships
+        # as an artifact. Analyzing a downloaded core against the runtime
+        # artifact's binary in an ubuntu-24.04 userspace reproduces the exact
+        # crash stack without burning a CI round on an in-workflow debugger.
+        env:
+          MOBILEGL_ITEST_REQUIRE_GPU: "1"
+          MOBILEGL_FIX_ITERATIONRP_SUBGROUP_SCRATCH: "1"
+          MOBILEGL_DERIVE_NUM_SUBGROUPS: "1"
+          MOBILEGL_ITERATIONRP_FIX_BARRIER: "1"
+        run: |
+          ulimit -c unlimited
+          sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
+          if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" = "true" ]; then
+            ctest -V -L integration-gpu --no-tests=error
+          else
+            ctest --output-on-failure -L integration-gpu --no-tests=error
+          fi
+
+      - name: Upload core dumps
+        if: failure()
+        uses: actions/upload-artifact@v7
+        with:
+          name: integration-core-dumps
+          path: /tmp/core.*
+          if-no-files-found: ignore
+
   benchmark:
     runs-on: ubuntu-latest
     needs: build-linux
@@ -177,9 +297,13 @@ jobs:
         uses: lukka/get-cmake@v4.3.3
 
       - name: Install runtime dependencies
+        # libegl-mesa0 is the EGL vendor library itself: DriverBench brings up a
+        # real GL context, and libegl1 is only glvnd's dispatch. It normally
+        # arrives as a Recommends of libegl1, which is too quiet a dependency for
+        # the one job that needs a working driver.
         run: |
           sudo apt-get update
-          sudo apt-get install -y libvulkan1 libegl1 libgles2 libgl1-mesa-dri mesa-vulkan-drivers
+          sudo apt-get install -y libvulkan1 libegl1 libegl-mesa0 libgles2 libgl1-mesa-dri mesa-vulkan-drivers
 
       - name: Download Linux runtime
         uses: actions/download-artifact@v8
@@ -204,7 +328,18 @@ jobs:
 
       - name: Benchmark
         working-directory: build-linux
-        run: ctest -V -C Release -L benchmark --no-tests=error
+        run: |
+          ulimit -c unlimited
+          sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
+          ctest -V -C Release -L benchmark --no-tests=error
+
+      - name: Upload core dumps
+        if: failure()
+        uses: actions/upload-artifact@v7
+        with:
+          name: benchmark-core-dumps
+          path: /tmp/core.*
+          if-no-files-found: ignore
 
   build-retrace:
     runs-on: ubuntu-latest
@@ -212,6 +347,10 @@ jobs:
       - build-linux
       - test
       - benchmark
+      - integration
+    permissions:
+      actions: write
+      contents: read
     env:
       BUILD_DIR: build-retrace
       CCACHE_BASEDIR: ${{ github.workspace }}
@@ -236,12 +375,11 @@ jobs:
         uses: lukka/get-cmake@v4.3.3
 
       - name: Restore ccache
-        uses: actions/cache@v5
+        uses: actions/cache/restore@v5
         with:
           path: .ccache
-          key: ${{ runner.os }}-test-${{ github.job }}-ccache-${{ github.ref_name }}-${{ github.run_id }}
+          key: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
           restore-keys: |
-            ${{ runner.os }}-test-${{ github.job }}-ccache-${{ github.ref_name }}-
             ${{ runner.os }}-test-${{ github.job }}-ccache-
 
       - name: Prepare Vulkan SDK
@@ -307,6 +445,21 @@ jobs:
         if: always()
         run: ccache --show-stats
 
+      - name: Release superseded ccache entry
+        if: github.ref_name == github.event.repository.default_branch
+        env:
+          GH_TOKEN: ${{ github.token }}
+          CACHE_KEY: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
+        run: gh cache delete "${CACHE_KEY}" || true
+
+      - name: Save ccache
+        if: github.ref_name == github.event.repository.default_branch
+        continue-on-error: true
+        uses: actions/cache/save@v5
+        with:
+          path: .ccache
+          key: ${{ runner.os }}-test-${{ github.job }}-ccache-v1
+
       - name: Normalize CTest command paths
         run: |
           python - <<'PY'
@@ -339,7 +492,9 @@ jobs:
     needs:
       - test
       - benchmark
+      - integration
     outputs:
+      matrix: ${{ steps.trace-cases.outputs.matrix }}
       names: ${{ steps.trace-cases.outputs.names }}
     steps:
       - name: Checkout repo
@@ -347,7 +502,9 @@ jobs:
 
       - name: Load trace cases
         id: trace-cases
-        run: echo "names=$(python3 tools/trace_replay/trace_cases.py --ci --format names)" >> "$GITHUB_OUTPUT"
+        run: |
+          echo "matrix=$(python3 tools/trace_replay/trace_cases.py --ci --format github-test-matrix)" >> "$GITHUB_OUTPUT"
+          echo "names=$(python3 tools/trace_replay/trace_cases.py --ci --format names)" >> "$GITHUB_OUTPUT"
 
   trace-fixtures:
     name: trace fixture (${{ matrix.case }})
@@ -362,9 +519,41 @@ jobs:
       - name: Checkout repo
         uses: actions/checkout@v6
 
+      - name: Derive trace fixture cache key
+        id: fixture-key
+        run: bash .github/scripts/trace-fixture-cache.sh key '${{ matrix.case }}'
+
+      - name: Restore trace fixture cache
+        id: fixture-cache
+        if: steps.fixture-key.outputs.cacheable == 'true'
+        uses: actions/cache/restore@v5
+        with:
+          path: ${{ steps.fixture-key.outputs.paths }}
+          key: ${{ steps.fixture-key.outputs.key }}
+
+      - name: Verify restored trace fixture
+        id: fixture-verify
+        if: steps.fixture-cache.outputs.cache-hit == 'true'
+        run: |
+          if bash .github/scripts/trace-fixture-cache.sh verify '${{ matrix.case }}'; then
+            echo "ok=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "ok=false" >> "$GITHUB_OUTPUT"
+            echo "::warning::Cached fixture for ${{ matrix.case }} failed verification; falling back to the download path"
+            bash .github/scripts/trace-fixture-cache.sh reset '${{ matrix.case }}'
+          fi
+
       - name: Fetch trace fixture
+        if: steps.fixture-verify.outputs.ok != 'true'
         run: bash .github/scripts/fetch-trace-fixture-lfs.sh '${{ matrix.case }}'
 
+      - name: Save trace fixture cache
+        if: steps.fixture-key.outputs.cacheable == 'true' && steps.fixture-cache.outputs.cache-hit != 'true'
+        uses: actions/cache/save@v5
+        with:
+          path: ${{ steps.fixture-key.outputs.paths }}
+          key: ${{ steps.fixture-key.outputs.key }}
+
       - name: Stage trace fixture
         run: |
           safe_case="$(printf '%s' '${{ matrix.case }}' | sed 's/[^A-Za-z0-9._-]/_/g')"
@@ -394,11 +583,7 @@ jobs:
     strategy:
       fail-fast: false
       max-parallel: 4
-      matrix:
+      matrix: ${{ fromJSON(needs.trace-cases.outputs.matrix) }}
-        backend:
-          - DirectGLES
-          - DirectVulkan
-        case: ${{ fromJSON(needs.trace-cases.outputs.names) }}
 
     steps:
       - name: Set Swap Space
@@ -452,9 +637,17 @@ jobs:
       - name: Retrace and validate
         working-directory: build-retrace/tools/trace_replay
         run: |
+          ulimit -c unlimited
+          sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
           if [ '${{ matrix.backend }}' = 'DirectVulkan' ]; then
             export MOBILEGL_MAGMA_R11G11B10F_FALLBACK=1
           fi
+          if [ '${{ matrix.backend }}' = 'DirectVulkan' ] \
+             && [ '${{ matrix.case }}' = 'minecraft-1.21.4-fabric-iris-iterationrp-in-world' ]; then
+            export MOBILEGL_FIX_ITERATIONRP_SUBGROUP_SCRATCH=1
+            export MOBILEGL_DERIVE_NUM_SUBGROUPS=1
+            export MOBILEGL_ITERATIONRP_FIX_BARRIER=1
+          fi
           # The blended depth-write quirk auto-enables only on Qualcomm, which no CI
           # runner has, so force it on for the OIT case it exists to fix. ForceOn
           # bypasses only the vendor gate, so this exercises the real strip on
@@ -466,6 +659,14 @@ jobs:
           fi
           ctest -V --no-tests=error -R '^MobileGLTraceReplay\.${{ matrix.case }}\.${{ matrix.backend }}$'
 
+      - name: Upload core dumps
+        if: failure()
+        uses: actions/upload-artifact@v7
+        with:
+          name: retrace-core-dumps-${{ matrix.backend }}-${{ matrix.case }}
+          path: /tmp/core.*
+          if-no-files-found: ignore
+
       - name: Upload actual image
         if: always()
         uses: actions/upload-artifact@v7

.gitignore

@@ -27,3 +27,4 @@ MobileGL/MG*/cmake-build*
 tools/trace_replay/work/
 __pycache__/
 *.py[cod]
+/.gradle

.gitmodules

@@ -7,9 +7,6 @@
 [submodule "3rdparty/SPIRV-Cross"]
 	path = 3rdparty/SPIRV-Cross
 	url = https://github.com/KhronosGroup/SPIRV-Cross.git
-[submodule "include/FastSTL"]
-	path = include/FastSTL
-	url = https://github.com/MobileGL-Dev/FastSTL.git
 [submodule "3rdparty/tracy"]
 	path = 3rdparty/tracy
 	url = https://github.com/wolfpld/tracy.git
@@ -31,3 +28,9 @@
 [submodule "3rdparty/apitrace"]
 	path = 3rdparty/apitrace
 	url = https://github.com/MobileGL-Dev/apitrace.git
+[submodule "3rdparty/asio"]
+	path = 3rdparty/asio
+	url = https://github.com/chriskohlhoff/asio.git
+[submodule "include/ska"]
+	path = include/ska
+	url = https://github.com/MobileGL-Dev/flat_hash_map.git

CMakeLists.txt

@@ -4,6 +4,11 @@ project("MobileGL")
 
 option(MOBILEGL_BUILD_TEST        "Build MobileGL tests"                             ON )
 option(MOBILEGL_BUILD_BENCHMARK   "Build MobileGL benchmarks"                        ON )
+# Headless end-to-end GPU scenarios (MobileGL/MG_IntegrationTest). They need a
+# real GPU/ICD to do anything, so they are off by default for CI; every scenario
+# skips cleanly where there is none. Registered under the `integration-gpu`
+# ctest label so a run can select or exclude them.
+option(MOBILEGL_BUILD_INTEGRATION_TEST "Build MobileGL headless GPU integration tests" OFF)
 option(MOBILEGL_FORCE_RELEASE_OPT "Enable Release optimization flags in Debug build" ON )
 option(MOBILEGL_ENABLE_TRACY      "Enable tracy for profiling"                       OFF)
 option(MOBILEGL_BUILD_TRACE_REPLAY "Build desktop apitrace replay runner"            OFF)
@@ -15,9 +20,86 @@ set(MOBILEGL_VULKAN_LIBRARY "" CACHE FILEPATH "Vulkan loader/MoltenVK library to
 if (ANDROID)
     set(MOBILEGL_BUILD_TEST       OFF CACHE BOOL "Build MobileGL tests"      FORCE)
     set(MOBILEGL_BUILD_BENCHMARK  OFF CACHE BOOL "Build MobileGL benchmarks" FORCE)
+
+    # ------- Android API level policy: minimum 26, decided here and only here -------
+    # MobileGL ships against API 26: the codebase must not use any API introduced
+    # after 26. That usage constraint is enforced where it is real - the shipping
+    # gradle build compiles at minSdk 26, where a newer API is simply undeclared
+    # and fails to compile. Configuring at a HIGHER level is therefore allowed
+    # (nothing in the tree may rely on it), but a LOWER level would change the
+    # libc contract underneath the shipped library and is refused.
+    #
+    # This has to live at configure time because the level cannot be corrected
+    # from a source header. A `#define __ANDROID_API__ 26` in a common header
+    # only rewrites the macro for the bionic headers that happen to be included
+    # after it; any libc++ header pulled in earlier has already latched its
+    # feature macros at the real configure-time level. libc++ and bionic then
+    # disagree about which symbols exist - libc++ calls e.g.
+    # pthread_cond_clockwait while bionic, re-read at the lowered level, has
+    # hidden its declaration. MobileGL/Defines.h carried exactly that pin from
+    # the first commit until it was removed; this guard is what replaces it.
+    #
+    # Read the level back from the compiler target triple first. Its trailing
+    # number (aarch64-none-linux-android26) is precisely what clang turns into
+    # __ANDROID_API__, so it cannot disagree with the compile itself, and it is
+    # already past every NDK normalisation step - codename aliases, "latest",
+    # and per-ABI minimum pull-ups. ANDROID_PLATFORM_LEVEL is the fallback for
+    # generators/languages where the triple variable is not populated.
+    #
+    # Note CMAKE_SYSTEM_VERSION is deliberately NOT consulted: it holds the API
+    # level only under the NDK's newer toolchain path, and is a meaningless 1
+    # when ANDROID_USE_LEGACY_TOOLCHAIN_FILE is on (which is what AGP has been
+    # defaulting to). Reading it would fail every legacy-mode build.
+    set(MOBILEGL_ANDROID_API_LEVEL 26)
+
+    set(_mobilegl_android_api "")
+    foreach (_mobilegl_api_triple "${CMAKE_CXX_COMPILER_TARGET}"
+                                  "${CMAKE_C_COMPILER_TARGET}")
+        if (NOT _mobilegl_android_api AND
+            _mobilegl_api_triple MATCHES "-android([0-9]+)$")
+            set(_mobilegl_android_api "${CMAKE_MATCH_1}")
+        endif()
+    endforeach()
+
+    foreach (_mobilegl_api_var ANDROID_PLATFORM_LEVEL ANDROID_NATIVE_API_LEVEL
+                               ANDROID_PLATFORM)
+        if (NOT _mobilegl_android_api AND ${_mobilegl_api_var})
+            string(REGEX REPLACE "^android-" ""
+                   _mobilegl_android_api "${${_mobilegl_api_var}}")
+        endif()
+    endforeach()
+
+    if (NOT _mobilegl_android_api MATCHES "^[0-9]+$")
+        message(FATAL_ERROR
+            "MobileGL: could not determine the Android API level (got "
+            "\"${_mobilegl_android_api}\"). Configure with the NDK toolchain "
+            "file and -DANDROID_PLATFORM=android-${MOBILEGL_ANDROID_API_LEVEL}.")
+    elseif (_mobilegl_android_api LESS MOBILEGL_ANDROID_API_LEVEL)
+        message(FATAL_ERROR
+            "MobileGL requires at least Android API ${MOBILEGL_ANDROID_API_LEVEL}, "
+            "but this build resolved to API ${_mobilegl_android_api}.\n"
+            "Configure with -DANDROID_PLATFORM=android-${MOBILEGL_ANDROID_API_LEVEL} "
+            "(gradle builds get this from minSdk ${MOBILEGL_ANDROID_API_LEVEL}, so "
+            "check that minSdk instead of adding an override).")
+    elseif (_mobilegl_android_api GREATER MOBILEGL_ANDROID_API_LEVEL)
+        message(STATUS
+            "MobileGL: configuring at Android API ${_mobilegl_android_api} "
+            "(> shipping minimum ${MOBILEGL_ANDROID_API_LEVEL}). Allowed, but the "
+            "tree must not use post-${MOBILEGL_ANDROID_API_LEVEL} APIs - the "
+            "minSdk-${MOBILEGL_ANDROID_API_LEVEL} gradle build is the enforcing "
+            "compile.")
+    endif()
+
+    message(STATUS "MobileGL: Android API level ${_mobilegl_android_api}")
+
+    unset(_mobilegl_android_api)
+    unset(_mobilegl_api_var)
+    unset(_mobilegl_api_triple)
 endif()
 
-if (NOT CMAKE_BUILD_TYPE STREQUAL "Debug" OR MOBILEGL_FORCE_RELEASE_OPT)
+option(MOBILEGL_ENABLE_LTO "Build with ThinLTO/IPO" OFF)
+
+if ((NOT CMAKE_BUILD_TYPE STREQUAL "Debug" OR MOBILEGL_FORCE_RELEASE_OPT) AND MOBILEGL_ENABLE_LTO)
     # Check if ThinLTO or LTO is suppported
     include(CheckIPOSupported)
     include(CheckCCompilerFlag)
@@ -100,6 +182,7 @@ set(ENABLE_SPVREMAPPER         OFF CACHE BOOL "Enable SPVRemapper" FORCE)
 set(ENABLE_OPT                 ON  CACHE BOOL "Enable SPIRV-Tools opt usage in glslang" FORCE)
 set(BUILD_EXTERNAL             ON  CACHE BOOL "Build external deps in External/" FORCE)
 set(ENABLE_GLSLANG_INSTALL     OFF CACHE BOOL "Install glslang targets" FORCE)
+set(SPIRV_SKIP_EXECUTABLES     ON  CACHE BOOL "Skip building SPIRV-Tools executables" FORCE)
 
 set(SPIRV_CROSS_C_API          ON  CACHE BOOL "Enable C API" FORCE)
 set(SPIRV_CROSS_ENABLE_GLSL    ON  CACHE BOOL "Enable GLSL backend" FORCE)
@@ -147,6 +230,9 @@ set(SOURCE_FILES
 
     MobileGL/MG_Util/Debug/Log.cpp
 
+    MobileGL/MG_Util/Async/JobNode.cpp
+    MobileGL/MG_Util/Async/ShaderCompilePool.cpp
+
     MobileGL/MG_Util/Math/VectorTypes.cpp
     MobileGL/MG_Util/Metrics/TextureMetrics.cpp
 
@@ -180,6 +266,7 @@ set(SOURCE_FILES
 
     MobileGL/MG_Util/Classifiers/TextureEnumClassifier.cpp
 
+    MobileGL/MG_Util/ShaderTranspiler/CompileEnv.cpp
     MobileGL/MG_Util/ShaderTranspiler/ShaderCompiler.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpvcSession.cpp
     MobileGL/MG_Util/ShaderTranspiler/ShaderSourceProcessor.cpp
@@ -187,24 +274,41 @@ set(SOURCE_FILES
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/FlattenInterfaceStructPass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/EliminateFloatEqualsZeroPass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/RenameSamplerFunctionParameterPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/RenameBuiltinShadowingFunctionsPass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/DecomposeWorkgroupVec3Pass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/DecoratePositionInvariantPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/DemoteFloat64Pass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/LowerDrawParametersPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/PackDoubleVertexInputsPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/FlattenXfbInterfaceBlocksPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/SplitArrayVertexInputsPass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/RebaseInstanceIndexPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/ZeroBaseVertexPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/DeriveNumSubgroupsPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/FixIterationRPBarrierPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/FixIterationRPSubgroupScratchPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/EmulateSubgroupsPass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/NormalizeRectCoordinatesPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/Lower1DArrayImagesPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/BakeImageFormatsPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/PrivateToEntryLocalPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/StripUniformLocationsPass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/StripUboMemberRelaxedPrecisionPass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/StripNoPerspectivePass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/EmulateNoPerspectivePass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/LegalizeFragmentOutputIndexPass.cpp
 
     MobileGL/MG_Util/BackendLoaders/OpenGL/Loader.cpp
     MobileGL/MG_Util/BackendLoaders/Vulkan/Loader.cpp
 
     MobileGL/MG_Util/SelfTest/DriverPost.cpp
+    MobileGL/MG_Util/SelfTest/DriverPostIterationRPWitness.cpp
 
     MobileGL/MG_Util/Texture/PixelStoreProcessor.cpp
     MobileGL/MG_Util/Texture/TextureFormatProcessor.cpp
 
     MobileGL/MG_Impl/GLXImpl/Exporting/Definitions.cpp
+    MobileGL/MG_Impl/GLXImpl/GLXImpl.cpp
     MobileGL/MG_Impl/GLXImpl/LookUp/LookUp.cpp
 
     MobileGL/MG_Impl/EGLImpl/Exporting/Definitions.cpp
@@ -218,6 +322,7 @@ set(SOURCE_FILES
     MobileGL/MG_Impl/GLImpl/Framebuffer/Validators.cpp
     MobileGL/MG_Impl/GLImpl/Framebuffer/GL_Framebuffer.cpp
     MobileGL/MG_Impl/GLImpl/Program/GL_Program.cpp
+    MobileGL/MG_Impl/GLImpl/Program/ProgramInterface.cpp
     MobileGL/MG_Impl/GLImpl/Program/GL_ProgramPipeline.cpp
     MobileGL/MG_Impl/GLImpl/Texture/GL_Texture.cpp
     MobileGL/MG_Impl/GLImpl/Texture/Validators.cpp
@@ -241,6 +346,7 @@ set(SOURCE_FILES
     MobileGL/MG_Backend/DirectGLES/BackendObject_DirectGLES.cpp
     MobileGL/MG_Backend/DirectGLES/Utils.cpp
     MobileGL/MG_Backend/DirectGLES/Managers.cpp
+    MobileGL/MG_Backend/DirectGLES/MultiDraw.cpp
 
     MobileGL/MG_Backend/DirectVulkan/DirectVulkan.cpp
     MobileGL/MG_Backend/DirectVulkan/BackendObject_DirectVulkan.cpp
@@ -279,7 +385,12 @@ set(SOURCE_FILES
     MobileGL/MG_State/GLState/TextureState/TextureUnit.cpp
     MobileGL/MG_State/GLState/TextureState/TextureState.cpp
     MobileGL/MG_State/GLState/ProgramState/ProgramObject.cpp
+    MobileGL/MG_State/GLState/ProgramState/ProgramLinkTask.cpp
+    MobileGL/MG_State/GLState/ProgramState/ProgramSpirvTask.cpp
+    MobileGL/MG_State/GLState/ProgramState/ShaderCompileTask.cpp
     MobileGL/MG_State/GLState/ProgramState/ShaderObject.cpp
+    MobileGL/MG_State/GLState/ProgramState/ShaderPreprocessCache.cpp
+    MobileGL/MG_State/GLState/ProgramState/ShaderCompileAdoptionMap.cpp
     MobileGL/MG_State/GLState/ProgramState/ProgramState.cpp
     MobileGL/MG_State/GLState/RenderState/RenderState.cpp
     MobileGL/MG_State/GLState/FramebufferState/FramebufferObject.cpp
@@ -302,6 +413,7 @@ endif()
 if (ANDROID)
     list(APPEND SOURCE_FILES
         MobileGL/MG_Util/SelfTest/DriverPostJni.cpp
+        MobileGL/MG_Util/SelfTest/DriverBenchJni.cpp
     )
 endif()
 
@@ -312,6 +424,11 @@ if (WIN32)
     )
 endif()
 
+# The shader-compile pool runs standalone Asio on real threads. This host's glibc (>= 2.34)
+# merged pthread into libc, so it links without asking, but the NDK and musl are not
+# guaranteed to be as forgiving - ask for it explicitly rather than rely on the accident.
+find_package(Threads REQUIRED)
+
 set(MOBILEGL_LINK_LIBRARIES
         glslang::glslang
         spirv-cross-c
@@ -321,12 +438,17 @@ set(MOBILEGL_LINK_LIBRARIES
         GPUOpen::VulkanMemoryAllocator
         Vulkan::UtilityHeaders
         spirv-reflect-static
+        Threads::Threads
 )
 
 set(MOBILEGL_COMPILE_DEF
         -DVMA_STATIC_VULKAN_FUNCTIONS=0
         -DVMA_DYNAMIC_VULKAN_FUNCTIONS=1
         -DVMA_VULKAN_VERSION=1001000
+        # Header-only Asio, no Boost, no deprecated interfaces. Set on the definition list
+        # rather than per-target so the shared library and the _s static target agree.
+        -DASIO_STANDALONE
+        -DASIO_NO_DEPRECATED
 )
 
 message(STATUS "MOBILEGL_COMPILE_DEF=${MOBILEGL_COMPILE_DEF}")
@@ -338,6 +460,10 @@ set(MOBILEGL_INCLUDE_DIR
         ${spirv-tools_SOURCE_DIR}/include
         ${spirv-tools_BINARY_DIR}
         ${SPIRV-Headers_SOURCE_DIR}/include
+        # Header-only submodule: no add_subdirectory, no link target. Only
+        # MG_Util/Async/ShaderCompilePool.cpp includes it, and it stays behind that file's
+        # pimpl so no consumer target needs this path.
+        ${CMAKE_SOURCE_DIR}/3rdparty/asio/include
 )
 
 add_library(${CMAKE_PROJECT_NAME} SHARED
@@ -535,6 +661,12 @@ if (NOT ANDROID)
         add_subdirectory(MobileGL/MG_Test)
     endif()
 
+    # After MG_Test so googletest is already available when the unit tests are
+    # built; the module fetches its own copy when they are not.
+    if (MOBILEGL_BUILD_INTEGRATION_TEST)
+        add_subdirectory(MobileGL/MG_IntegrationTest)
+    endif()
+
     if (MOBILEGL_BUILD_BENCHMARK)
         add_subdirectory(MobileGL/MG_Benchmark)
     endif()
@@ -543,3 +675,10 @@ if (NOT ANDROID)
         add_subdirectory(tools/trace_replay)
     endif()
 endif()
+
+# The integration binary is also useful as a standalone adb-shell executable.
+# Android cannot use the desktop-only MobileGL_s target, so its CMake module
+# links libMobileGL.so and creates an AImageReader-backed window instead.
+if (ANDROID AND MOBILEGL_BUILD_INTEGRATION_TEST)
+    add_subdirectory(MobileGL/MG_IntegrationTest)
+endif()

MobileGL/Config.h

@@ -29,6 +29,33 @@ namespace MobileGL::MG_Config {
         ForceOff,
     };
 
+    // Preferred DirectVulkan dispatch tier for the glMultiDraw* families. A preference,
+    // never a demand: the renderer clamps it to what the device supports at device
+    // creation, falling down the chain ext -> indirect -> unroll with one log line.
+    enum class MultiDrawMode : Uint8 {
+        Auto = 0, // unset: best supported tier
+        Ext,      // VK_EXT_multi_draw: one vkCmdDrawMultiEXT / vkCmdDrawMultiIndexedEXT
+        Indirect, // multiDrawIndirect feature: one vkCmdDraw*Indirect over a transient command array
+        Unroll,   // one vkCmdDraw* per sub-draw
+    };
+
+    // Preferred DirectGLES emulation tier for glMultiDrawElements(BaseVertex). GLES has no
+    // such entry point in core, so every tier below is an emulation; they differ only in
+    // which driver capability they lean on and how many driver calls a batch costs. Like
+    // the Magma knob this is a preference, clamped at resolution time to what the ES
+    // driver actually supports, with one log line when it falls back.
+    enum class GLESMultiDrawMode : Uint8 {
+        Auto = 0,      // unset: best supported tier
+        Ext,           // one glMultiDrawElementsBaseVertexEXT
+        MultiIndirect, // one glMultiDrawElementsIndirectEXT over a scratch command buffer
+        Indirect,      // one glDrawElementsIndirect per sub-draw over that same buffer
+        BaseVertex,    // one glDrawElementsBaseVertex per sub-draw
+        DrawElements,  // baseVertex folded into a scratch index buffer on the CPU, then plain
+                       // glDrawElements per sub-draw (for drivers with no base-vertex draw at all)
+        Compute,       // a compute shader flattens every sub-draw into one rebased index buffer,
+                       // drawn by a single glDrawElements
+    };
+
     // Feature toggles parsed once from environment variables in MG_ConfigLoader::Init()
     // (ConfigLoader.cpp), before the accepted-env map is destroyed. All Bool fields share
     // one truthy rule: the variable is set, non-empty, not "0", and not "false"
@@ -42,14 +69,58 @@ namespace MobileGL::MG_Config {
     struct FeaturesTable {
         // MOBILEGL_DISABLE_TIMERQUERY: do not advertise or use GPU timer queries.
         Bool DisableTimerQuery = false;
+        // MOBILEGL_ENABLE_SPIRV_VALIDATION: validate generated and transformed SPIR-V.
+        // Disabled by default because validation is a diagnostics-only cost.
+        Bool EnableSpirvValidation = false;
         // MOBILEGL_USE_ANGLE: load ANGLE EGL/GLES libraries.
         Bool UseAngle = false;
 #if defined(MOBILEGL_TRACE_ANGLE_VARIANTS)
         // MOBILEGL_TRACE_ANGLE_VARIANT: signed trace-APK ANGLE build short hash.
         String TraceAngleVariant;
 #endif
-        // MOBILEGL_DISABLE_SUBGROUP: force-disable Vulkan shader subgroup support.
+        // MOBILEGL_DISABLE_SUBGROUP: force-disable Vulkan shader subgroup support,
+        // including the opt-in emulated compute path below.
         Bool DisableSubgroup = false;
+        // MOBILEGL_MAGMA_EMULATE_SUBGROUP: implement GL_KHR_shader_subgroup's compute
+        // stage on a 32-lane VIRTUAL subgroup lowered to workgroup-shared memory
+        // (ShaderTranspiler::EmulateSubgroupsPass). Strictly a last resort: it only ever
+        // engages when this flag is set AND the device has no native subgroup support at
+        // all - a device with real subgroup operations always uses them natively,
+        // whatever their width (the known iterationRP defect is patched by
+        // FixIterationRPSubgroupScratch below instead). Off by default.
+        Bool MagmaEmulateSubgroup = false;
+        // MOBILEGL_FIX_ITERATIONRP_SUBGROUP_SCRATCH: patch iterationRP's own bug - the
+        // pack declares `shared vec2 prefixSumCache[32]` for a 512-invocation exposure
+        // reduction and indexes it by gl_SubgroupID, so any device with sub-16-lane
+        // subgroups (8-lane lavapipe -> 64 subgroups) writes shared memory out of
+        // bounds. The pass grows that one array to what the device's topology needs and
+        // touches nothing else; it only rewrites modules positively matching the pack's
+        // reduction fingerprint (ShaderTranspiler::FixIterationRPSubgroupScratchPass),
+        // so every other shader passes through byte-identical - as does iterationRP
+        // itself on >= 16-lane devices. Auto is ON; ForceOff replays the pack's bug
+        // verbatim.
+        QuirkOverride FixIterationRPSubgroupScratch = QuirkOverride::Auto;
+        // MOBILEGL_ITERATIONRP_FIX_BARRIER: repair Program 203's missing workgroup
+        // rendezvous between its two reductions over prefixSumCache. Off by default and
+        // fingerprint-gated by FixIterationRPBarrierPass when enabled.
+        Bool IterationRPFixBarrier = false;
+        // MOBILEGL_DERIVE_NUM_SUBGROUPS: replace compute gl_NumSubgroups loads with
+        // ceil(workgroup invocations / gl_SubgroupSize) on the NATIVE subgroup path
+        // (ShaderTranspiler::DeriveNumSubgroupsPass). Auto is ON: GL requires
+        // gl_SubgroupID < gl_NumSubgroups, Adreno's builtin reports 1 while the same
+        // dispatch emits IDs 0..7, and the derived value is the one Vulkan guarantees
+        // whenever the pipeline can request REQUIRE_FULL_SUBGROUPS (which the renderer
+        // does whenever local_size_x is a multiple of the native width). ForceOff returns
+        // to the raw driver builtin.
+        QuirkOverride DeriveNumSubgroups = QuirkOverride::Auto;
+        // MOBILEGL_ADVERTISE_FP64: add GL_ARB_gpu_shader_fp64 to the advertised extension
+        // string. `double` in a shader always WORKS - it is narrowed to 32 bits before any
+        // module reaches a backend (ShaderTranspiler::DemoteFloat64Pass) - but the extension
+        // promises 64-bit precision, and that is the one thing the narrowing cannot deliver.
+        // Off by default so an application that checks the string before using doubles keeps
+        // its float path; on for measuring what the conformance suite makes of the demoted
+        // precision. See the DemoteFloat64Pass header and the "fp64" POST row.
+        Bool AdvertiseFp64 = false;
         // MOBILEGL_MAGMA_R11G11B10F_FALLBACK: use fallback format for R11G11B10F on Vulkan.
         Bool MagmaR11G11B10FFallback = false;
         // MOBILEGL_MAGMA_FRAMESINFLIGHT: requested Magma frames in flight, defaulting to 3.
@@ -57,6 +128,13 @@ namespace MobileGL::MG_Config {
         // MOBILEGL_AVOID_SAMPLER_MIPMAP_MIN_FILTER: avoid mipmap min filters in samplers,
         // resolves certain rendering bugs on ANGLE + llvmpipe.
         Bool AvoidSamplerMipmapMinFilter = false;
+        // MOBILEGL_AVOID_EXPLICIT_LOD_BIAS: leave an already-explicit LOD argument alone when
+        // emulating GL_TEXTURE_LOD_BIAS, instead of adding the bias uniform to it. Injecting
+        // the uniform turns a compile-time-constant LOD into a runtime expression, which
+        // sends ANGLE + llvmpipe down a mip-selection path that dereferences a NULL
+        // descriptor and kills the process. Deviates from spec (Vulkan adds the bias to
+        // OpImageSampleExplicitLod), so it is an avoidance for that stack only.
+        Bool AvoidExplicitLodBias = false;
         // MOBILEGL_COHERENT_AS_FLUSH: app-compat for engines (e.g. Flywheel) that write
         // GPU-read data through persistent GL_MAP_FLUSH_EXPLICIT_BIT maps they never
         // flush. Persistent FLUSH_EXPLICIT map requests are rewritten to coherent
@@ -70,16 +148,19 @@ namespace MobileGL::MG_Config {
         // per-draw glBufferSubData path instead of the persistent-mapped ring allocator
         // (negative control / driver-bug escape hatch).
         Bool DisableUboRing = false;
+        // MOBILEGL_ESPRYT_FORCE_DS_READBACK_EMULATION: make DirectGLES skip the native ES
+        // depth/stencil reads and always go through the shader-sampling emulation. Core GL
+        // ES has no depth or stencil readback, but some drivers accept it anyway (Mesa does,
+        // Adreno does not), which means the emulation is dead code on exactly the stack the
+        // headless suite runs on. This forces it live so the scenarios and the CTS can
+        // exercise the path, and gives the device an A/B lever over the same choice.
+        Bool EsprytForceDepthStencilReadbackEmulation = false;
         // MOBILEGL_RELAXED_SEMANTICS: relax strict core-profile rules (e.g. VAO-0 draws,
         // texture-name reuse after delete) even on contexts that explicitly requested a core
         // profile. Without it, relaxed semantics still apply to every context that did not
         // explicitly request a core profile via EGL_CONTEXT_OPENGL_PROFILE_MASK / a >=3.1
         // version request.
         Bool RelaxedSemantics = false;
-        // MOBILEGL_QUIRK_SUBGROUP_PREFIX_SCAN: overrides the shader-source quirk that
-        // rewrites the recognized workgroup prefix-scan template on Qualcomm devices with
-        // subgroups wider than 32 lanes (see ShaderSourceProcessor's quirk registry).
-        QuirkOverride SubgroupPrefixScanQuirk = QuirkOverride::Auto;
         // MOBILEGL_MAGMA_DISABLE_BLENDED_DEPTH_WRITE: overrides the DirectVulkan quirk that
         // strips depth writes from accumulation-blended pipelines (MIN/MAX or additive
         // ONE+ONE - the multi-pass depth-equality signature) on drivers without
@@ -91,6 +172,37 @@ namespace MobileGL::MG_Config {
         // feature off. It is enabled by default to match GL's defined out-of-range fetch
         // behavior; this escape hatch exists to measure or dodge its GPU cost on a device.
         Bool DisableRobustBufferAccess = false;
+        // MOBILEGL_MAGMA_MULTIDRAW_MODE: preferred DirectVulkan multi-draw dispatch tier
+        // ("ext" | "indirect" | "unroll", see MultiDrawMode). Clamped to device support;
+        // unset picks the best supported tier.
+        MultiDrawMode MagmaMultiDrawMode = MultiDrawMode::Auto;
+        // MOBILEGL_ESPRYT_MULTIDRAW_MODE: preferred DirectGLES glMultiDrawElements emulation
+        // tier ("ext" | "multiindirect" | "indirect" | "basevertex" | "drawelements" |
+        // "compute", see GLESMultiDrawMode). Clamped to driver support; unset picks the best
+        // supported tier, which never includes "compute" - see the note on its resolution.
+        GLESMultiDrawMode EsprytMultiDrawMode = GLESMultiDrawMode::Auto;
+        // MOBILEGL_ASYNC_SHADER_COMPILE: overrides asynchronous shader compilation. Unset
+        // keeps the built-in default (MG_Util::Async::kAsyncShaderCompileDefault); falsy
+        // forces every glCompileShader/glLinkProgram to run synchronously on the calling
+        // thread AND withdraws GL_KHR_parallel_shader_compile, so the single switch reverts
+        // both the threading and the application-visible behaviour change.
+        QuirkOverride AsyncShaderCompile = QuirkOverride::Auto;
+        // MOBILEGL_ASYNC_SHADER_COMPILE_THREADS: shader-compile worker count. 0 (unset) means
+        // auto, which is min(4, big cores); an explicit value is honoured as given.
+        Uint32 AsyncShaderCompileThreads = 0;
+        // MOBILEGL_ASYNC_OPTIMISTIC_SHADER_STATUS: while a compile job is still in flight,
+        // glGetShaderiv(GL_COMPILE_STATUS) answers GL_TRUE and the shader info log reads
+        // empty, WITHOUT joining the job (latched per compile - see
+        // ShaderObject::TakeOptimisticCompileAnswer). A deliberate, bounded spec violation:
+        // a real failure still fails the program link with the compile log quoted. It
+        // exists for applications that compile hundreds of shaders serially and read the
+        // status right after each glCompileShader - Iris's shader-pack load - where those
+        // per-shader joins are what serializes the batch on its main path (Iris's gbuffer
+        // phase issues no program-level query between programs; program-level LINK_STATUS
+        // and the program info log still join truthfully, so paths that check each link
+        // immediately stay serial by their own construction). Off by default; never
+        // advertise it.
+        QuirkOverride AsyncOptimisticShaderStatus = QuirkOverride::Auto;
     };
     extern FeaturesTable Features;
 } // namespace MobileGL::MG_Config

MobileGL/ConfigLoader.cpp

@@ -97,6 +97,47 @@ namespace MobileGL::MG_ConfigLoader {
                                          : MG_Config::QuirkOverride::ForceOff;
     }
 
+    // Multi-draw mode is a named-value preference: unset keeps Auto (best supported tier),
+    // a recognized name selects that tier as the ceiling, anything else warns and keeps Auto.
+    inline MG_Config::MultiDrawMode QueryEnvMultiDrawMode(const String& key) {
+        auto it = acceptedEnvVariablesMap->find(key);
+        if (it == acceptedEnvVariablesMap->end()) {
+            return MG_Config::MultiDrawMode::Auto;
+        }
+        String lowered = it->second;
+        std::transform(lowered.begin(), lowered.end(), lowered.begin(),
+                       [](unsigned char c) { return static_cast<char>(std::tolower(c)); });
+        if (lowered == "ext") return MG_Config::MultiDrawMode::Ext;
+        if (lowered == "indirect") return MG_Config::MultiDrawMode::Indirect;
+        if (lowered == "unroll") return MG_Config::MultiDrawMode::Unroll;
+        if (lowered.empty() || lowered == "auto") return MG_Config::MultiDrawMode::Auto;
+        MGLOG_W("Config: Ignoring invalid env variable %s='%s'; expected ext|indirect|unroll|auto, using auto",
+                key.c_str(), it->second.c_str());
+        return MG_Config::MultiDrawMode::Auto;
+    }
+
+    // Same contract as QueryEnvMultiDrawMode, over the DirectGLES tier names.
+    inline MG_Config::GLESMultiDrawMode QueryEnvGLESMultiDrawMode(const String& key) {
+        auto it = acceptedEnvVariablesMap->find(key);
+        if (it == acceptedEnvVariablesMap->end()) {
+            return MG_Config::GLESMultiDrawMode::Auto;
+        }
+        String lowered = it->second;
+        std::transform(lowered.begin(), lowered.end(), lowered.begin(),
+                       [](unsigned char c) { return static_cast<char>(std::tolower(c)); });
+        if (lowered == "ext") return MG_Config::GLESMultiDrawMode::Ext;
+        if (lowered == "multiindirect") return MG_Config::GLESMultiDrawMode::MultiIndirect;
+        if (lowered == "indirect") return MG_Config::GLESMultiDrawMode::Indirect;
+        if (lowered == "basevertex") return MG_Config::GLESMultiDrawMode::BaseVertex;
+        if (lowered == "drawelements") return MG_Config::GLESMultiDrawMode::DrawElements;
+        if (lowered == "compute") return MG_Config::GLESMultiDrawMode::Compute;
+        if (lowered.empty() || lowered == "auto") return MG_Config::GLESMultiDrawMode::Auto;
+        MGLOG_W("Config: Ignoring invalid env variable %s='%s'; expected "
+                "ext|multiindirect|indirect|basevertex|drawelements|compute|auto, using auto",
+                key.c_str(), it->second.c_str());
+        return MG_Config::GLESMultiDrawMode::Auto;
+    }
+
     inline Uint32 QueryEnvUint32(const String& key, Uint32 defaultValue, Uint32 minValue, Uint32 maxValue) {
         auto it = acceptedEnvVariablesMap->find(key);
         if (it == acceptedEnvVariablesMap->end()) {
@@ -121,23 +162,38 @@ namespace MobileGL::MG_ConfigLoader {
     inline void InitFeatures() {
         auto& features = MG_Config::Features;
         features.DisableTimerQuery = QueryEnvFlag("MOBILEGL_DISABLE_TIMERQUERY");
+        features.EnableSpirvValidation = QueryEnvFlag("MOBILEGL_ENABLE_SPIRV_VALIDATION");
         features.UseAngle = QueryEnvFlag("MOBILEGL_USE_ANGLE");
 #if defined(MOBILEGL_TRACE_ANGLE_VARIANTS)
         QueryEnvVariable("MOBILEGL_TRACE_ANGLE_VARIANT", features.TraceAngleVariant, "");
 #endif
         features.DisableSubgroup = QueryEnvFlag("MOBILEGL_DISABLE_SUBGROUP");
+        features.MagmaEmulateSubgroup = QueryEnvFlag("MOBILEGL_MAGMA_EMULATE_SUBGROUP");
+        features.FixIterationRPSubgroupScratch =
+            QueryEnvQuirkOverride("MOBILEGL_FIX_ITERATIONRP_SUBGROUP_SCRATCH");
+        features.IterationRPFixBarrier = QueryEnvFlag("MOBILEGL_ITERATIONRP_FIX_BARRIER");
+        features.DeriveNumSubgroups = QueryEnvQuirkOverride("MOBILEGL_DERIVE_NUM_SUBGROUPS");
+        features.AdvertiseFp64 = QueryEnvFlag("MOBILEGL_ADVERTISE_FP64");
         features.MagmaR11G11B10FFallback = QueryEnvFlag("MOBILEGL_MAGMA_R11G11B10F_FALLBACK");
         features.MagmaFramesInFlight = QueryEnvUint32("MOBILEGL_MAGMA_FRAMESINFLIGHT", 3, 1, 64);
         features.AvoidSamplerMipmapMinFilter =
             QueryEnvFlag("MOBILEGL_AVOID_SAMPLER_MIPMAP_MIN_FILTER");
+        features.AvoidExplicitLodBias = QueryEnvFlag("MOBILEGL_AVOID_EXPLICIT_LOD_BIAS");
         features.CoherentAsFlush = QueryEnvFlag("MOBILEGL_COHERENT_AS_FLUSH");
         features.TraceSkipAutodestroy = QueryEnvFlag("MOBILEGL_TRACE_SKIP_AUTODESTROY");
         features.DisableUboRing = QueryEnvFlag("MOBILEGL_DISABLE_UBO_RING");
+        features.EsprytForceDepthStencilReadbackEmulation =
+            QueryEnvFlag("MOBILEGL_ESPRYT_FORCE_DS_READBACK_EMULATION");
         features.RelaxedSemantics = QueryEnvFlag("MOBILEGL_RELAXED_SEMANTICS");
-        features.SubgroupPrefixScanQuirk = QueryEnvQuirkOverride("MOBILEGL_QUIRK_SUBGROUP_PREFIX_SCAN");
         features.MagmaDisableBlendedDepthWriteQuirk =
             QueryEnvQuirkOverride("MOBILEGL_MAGMA_DISABLE_BLENDED_DEPTH_WRITE");
         features.DisableRobustBufferAccess = QueryEnvFlag("MOBILEGL_DISABLE_ROBUST_BUFFER_ACCESS");
+        features.MagmaMultiDrawMode = QueryEnvMultiDrawMode("MOBILEGL_MAGMA_MULTIDRAW_MODE");
+        features.EsprytMultiDrawMode = QueryEnvGLESMultiDrawMode("MOBILEGL_ESPRYT_MULTIDRAW_MODE");
+        features.AsyncShaderCompile = QueryEnvQuirkOverride("MOBILEGL_ASYNC_SHADER_COMPILE");
+        features.AsyncShaderCompileThreads = QueryEnvUint32("MOBILEGL_ASYNC_SHADER_COMPILE_THREADS", 0, 0, 64);
+        features.AsyncOptimisticShaderStatus =
+            QueryEnvQuirkOverride("MOBILEGL_ASYNC_OPTIMISTIC_SHADER_STATUS");
     }
 
     inline void InitBackendType() {

MobileGL/Defines.h

@@ -9,10 +9,20 @@
 #pragma once
 
 // ============== Platform-specific definitions and macros ============== //
-#ifdef __ANDROID__
+// No __ANDROID_API__ pin here on purpose. The effective API level is owned by
-#undef __ANDROID_API__
+// the build system (gradle minSdk 26 -> -DANDROID_PLATFORM=android-26, enforced
-#define __ANDROID_API__ 26 // force Android API level to 26 for compatibility
+// by the configure-time guard in CMakeLists.txt), not by a macro.
-#endif
+//
+// History: this used to `#define __ANDROID_API__ 26` to *raise* the level back
+// when the build configured something lower, so that pthread_getname_np (which
+// bionic guards with __INTRODUCED_IN(26)) would be declared. Once a later
+// change added an `#undef` in front of it, the same line started *lowering* the
+// level whenever the build configured higher than 26 - and that is an
+// include-order split-brain, not a compatibility knob: a TU that includes any
+// libc++ header before Includes.h latches libc++'s feature macros at the
+// configure-time level, and only the bionic headers pulled in afterwards see
+// the lowered value. The two halves then disagree (e.g. libc++ believes
+// pthread_cond_clockwait exists while bionic has since hidden its declaration).
 
 #ifdef _WIN32
 #ifndef NOMINMAX
@@ -37,6 +47,23 @@
 #define MOBILEGL_WGL_API MOBILEGL_API
 
 // ====================== MobileGL configurations ======================= //
+// The numeric log levels live here, not only in Log.h: MOBILEGL_ASSERT below compares
+// MOBILEGL_LOG_ACTIVE_LEVEL against MOBILEGL_LOG_LEVEL_DEBUG, and in a translation unit
+// that includes Defines.h without Log.h both tokens would silently evaluate to 0 in the
+// preprocessor conditional - enabling the assert in exactly the INFO-level builds it is
+// documented to be compiled out of. Log.h redefines them identically, which is legal.
+//
+// Severity order, ascending: DEBUG < INFO < WARN < ERROR < FATAL. MOBILEGL_LOG_ACTIVE_LEVEL
+// names the lowest severity compiled in, so the production default INFO keeps I/W/E/F and
+// drops only D. Any edit here must be mirrored in Log.h.
+#ifndef MOBILEGL_LOG_LEVEL_DEBUG
+#define MOBILEGL_LOG_LEVEL_DEBUG 0
+#define MOBILEGL_LOG_LEVEL_INFO 1
+#define MOBILEGL_LOG_LEVEL_WARN 2
+#define MOBILEGL_LOG_LEVEL_ERROR 3
+#define MOBILEGL_LOG_LEVEL_FATAL 4
+#endif
+
 #ifndef MOBILEGL_LOG_ACTIVE_LEVEL
 #define MOBILEGL_LOG_ACTIVE_LEVEL MOBILEGL_LOG_LEVEL_INFO
 #endif
@@ -68,6 +95,12 @@
 #endif
 
 // =============================== Utils ================================ //
+// Asserts are live in exactly the builds where MGLOG_D is live, i.e. DEBUG builds only;
+// an INFO build (the production default) compiles them out. DEBUG is the lowest severity
+// in the ordering above, so "ACTIVE <= DEBUG" is true only for ACTIVE == DEBUG - the same
+// gate MGLOG_D uses in Log.h. That equivalence is what makes this gate survive the
+// 2026-08-13 renumbering unchanged; the contract is and stays
+// "INFO builds: asserts OFF; DEBUG builds: asserts ON".
 #if MOBILEGL_LOG_ACTIVE_LEVEL <= MOBILEGL_LOG_LEVEL_DEBUG
     #define MOBILEGL_ASSERT(condition, ...)                                                                                \
         do {                                                                                                               \

MobileGL/Includes.h

@@ -49,8 +49,8 @@
 #include <stacktrace>
 #endif
 
-// Include FastSTL
+// Include ska::flat_hash_map
-#include <FastSTL/UnorderedMap.h>
+#include <ska/flat_hash_map.hpp>
 
 // Include xxHash
 #include <xxhash.h>

MobileGL/Init.cpp

@@ -15,6 +15,9 @@
 #include <MG_Impl/GLImpl/Texture/ProxyTexture.h>
 #include <MG_Impl/GLImpl/Framebuffer/GL_Framebuffer.h>
 #include <MG_Impl/GLImpl/Sync/GL_Sync.h>
+#include <MG_Impl/GLImpl/Query/GL_Query.h>
+#include <MG_Util/Async/ShaderCompilePool.h>
+#include <MG_Util/ShaderTranspiler/ShaderCompiler.h>
 
 #include <atomic>
 #include <mutex>
@@ -37,18 +40,38 @@ namespace MobileGL {
             if (logLifecycle) {
                 MGLOG_I("MobileGL closing...");
             }
-            glslang::FinalizeProcess();
+            // First, before anything else is torn down. In-flight compile/link jobs own
+            // their own inputs and are safe against everything below EXCEPT glslang's
+            // process globals and the TShader/TProgram objects hanging off pGLContext,
+            // both of which this function is about to destroy. This is the one
+            // cancellation path in the whole design that waits.
+            MG_Util::Async::ShaderCompilePool::Get().StopAndDrain();
             // GL syncs die with their contexts, and every context is gone by the
             // time full teardown runs: drain the live-sync registry while the
             // backend function table can still release the backend handles (and
             // before a re-initialized library could pair them with the wrong
             // backend's DeleteSync).
             MG_Impl::GLImpl::DestroyAllSyncObjects();
+            // Queries die with their contexts for the same reason, and their registry
+            // is the same shape of process-global map: drain it here too, while the
+            // function table can still pair each backend handle with the backend that
+            // minted it.
+            MG_Impl::GLImpl::DestroyAllQueryObjects();
             MG_Backend::pActiveBackendObject.reset();
             MG_State::pGLContext.reset();
             MG_State::pEGLContext.reset();
             MG_Impl::GLImpl::TextureImpl::pProxyTextureManager.reset();
             MG_Impl::GLImpl::FramebufferImpl::pDefaultFramebufferInfo.reset();
+            // Must run AFTER pGLContext.reset(). FinalizeProcess -> ShFinalize deletes
+            // glslang's process-wide pool allocator and every cached built-in symbol table,
+            // while the TShader/TProgram objects owned by the shader and program objects
+            // still reference levels adopted from those tables. Finalizing first left live
+            // glslang objects pointing at freed memory for the rest of the teardown.
+            glslang::FinalizeProcess();
+            // Immediately after, and never apart from it: FinalizeProcess just deleted the
+            // built-in symbol tables the prewarm latch stands for, so leaving it set would
+            // make the next Initialize() skip a prewarm it genuinely needs.
+            MG_Util::ShaderTranspiler::ShaderCompiler::ResetPrewarmLatch();
             MG_Backend::gBackendFunctionsTable = {};
             g_isInitialized = false;
             if (logLifecycle) {
@@ -76,6 +99,19 @@ namespace MobileGL {
         MG_Impl::Init();
         MGLOG_D("MG_Impl initialized");
         glslang::InitializeProcess();
+        // On the GL thread, before any worker can exist. glslang builds its built-in symbol
+        // tables lazily under a process-wide lock held for the whole build, so without this
+        // the first concurrent compiles of a shaderpack all serialize behind the very first
+        // parse and asynchronous compilation looks like it is doing nothing.
+        //
+        // Gated on the flag, because the problem it solves only exists when there are
+        // workers: with compilation synchronous, nothing ever contends for that lock and the
+        // three throwaway parses buy nothing - they just add to every eglInitialize. Read the
+        // flag here rather than inside PrewarmBuiltins so ShaderCompiler keeps no dependency
+        // on the async subsystem (ProgramUtilTest compiles that file without it).
+        if (MG_Util::Async::AsyncShaderCompileEnabled()) {
+            MG_Util::ShaderTranspiler::ShaderCompiler::PrewarmBuiltins();
+        }
         MGLOG_D("glslang initialized");
         g_isInitialized = true;
         MGLOG_I("MobileGL initialized");

MobileGL/MG_Backend/BackendObject.h

@@ -181,15 +181,18 @@ namespace MobileGL {
             void (*GetIntegeri_v)(GLenum target, GLuint index, GLint* data);
             void (*GetInteger64i_v)(GLenum target, GLuint index, GLint64* data);
             void (*GetProgramiv)(GLuint program, GLenum pname, GLint* params);
-            void (*GetProgramInterfaceiv)(GLuint program, GLenum programInterface, GLenum pname, GLint* params);
+            // The GL program interface (glGetProgramInterfaceiv / glGetProgramResource*) is NOT
-            GLuint (*GetProgramResourceIndex)(GLuint program, GLenum programInterface, const GLchar* name);
+            // a backend query: it describes the program the application wrote, in the
-            void (*GetProgramResourceName)(GLuint program, GLenum programInterface, GLuint index, GLsizei bufSize,
+            // application's namespace, which neither backend program is in. It is answered
-                                           GLsizei* length, GLchar* name);
+            // entirely by MG_Impl/GLImpl/Program/ProgramInterface from the frontend reflection.
-            void (*GetProgramResourceiv)(GLuint program, GLenum programInterface, GLuint index, GLsizei propCount,
+            // Takes the block's GL NAME, not glShaderStorageBlockBinding's index. The index
-                                         const GLenum* props, GLsizei bufSize, GLsizei* length, GLint* params);
+            // the application passes is the frontend interface-query enumeration's, and no
-            GLint (*GetProgramResourceLocation)(GLuint program, GLenum programInterface, const GLchar* name);
+            // backend shares that index space: DirectVulkan enumerates SPIR-V descriptor
-            GLint (*GetProgramResourceLocationIndex)(GLuint program, GLenum programInterface, const GLchar* name);
+            // bindings and DirectGLES asks a real driver about SPIRV-Cross-generated ESSL.
-            void (*ShaderStorageBlockBinding)(GLuint program, GLuint storageBlockIndex, GLuint storageBlockBinding);
+            // The name is the one coordinate all three agree on, so the frontend resolves the
+            // index against its own enumeration and each backend maps the name to its own.
+            void (*ShaderStorageBlockBinding)(GLuint program, const GLchar* storageBlockName,
+                                              GLuint storageBlockBinding);
             // GL fence sync objects. All entries are optional (may be null); the
             // frontend then falls back to always-signaled sync semantics.
             // FenceSync may itself return null when the backend cannot create a
@@ -358,7 +361,35 @@ namespace MobileGL {
             // glFramebufferTextureLayer, so it does; DirectVulkan maps a GL layer onto a Vulkan
             // array layer with no notion of a 3D depth slice, so it does not yet. Defaults to false
             // so a backend that never sets it gets the conservative answer.
-            Bool SupportsPerLayerFramebufferAttachment = false;
+            // Which layered texture targets this backend can attach ONE layer of to a framebuffer
+            // and then really clear, render and read back that layer. Bit (1u << TextureTarget) is
+            // set for each supported target. Deliberately per target rather than one flag: the three
+            // ways a GL layer maps onto Vulkan are independent capabilities. A 2D or 2D multisample
+            // array layer IS a VkImage array layer and needs nothing extra; a 3D texture's layer is
+            // a z slice, which needs a 2D-array-compatible image and a per-slice clear that
+            // vkCmdClearColorImage cannot express; a cube map array needs an image shape and the
+            // imageCubeArray feature before it can be attached at any layer at all. Defaults to 0 so
+            // a backend that never sets it gets the conservative answer.
+            Uint32 PerLayerFramebufferAttachmentTargets = 0;
+
+            static constexpr Uint32 PerLayerFramebufferAttachmentBit(TextureTarget target) {
+                return (static_cast<Int>(target) >= 0 &&
+                        static_cast<Int>(target) < static_cast<Int>(TextureTarget::TextureTargetCount))
+                           ? (1u << static_cast<Uint32>(target))
+                           : 0u;
+            }
+
+            Bool SupportsPerLayerFramebufferAttachment(TextureTarget target) const {
+                const Uint32 bit = PerLayerFramebufferAttachmentBit(target);
+                return bit != 0 && (PerLayerFramebufferAttachmentTargets & bit) != 0;
+            }
+            // Whether glVertexAttribLFormat / glVertexArrayAttribLFormat can be honoured, i.e.
+            // whether a 64-bit vertex attribute can actually reach a shader unconverted. Detected,
+            // never assumed: DirectVulkan needs VkPhysicalDeviceFeatures::shaderFloat64 (the
+            // attribute travels as its 32-bit word pair, so no VK_FORMAT_R64* is required, but the
+            // bitcast result is Float64); DirectGLES can never have it, ESSL having no fp64 type at
+            // all. Defaults to false so a backend that never sets it gets the conservative answer.
+            Bool SupportsFloat64VertexAttributes = false;
             SizeT MaxShaderStorageBlockSize = 128 * 1024 * 1024;
             Uint32 SubgroupSize = 0;
             Uint32 SubgroupSupportedStages = 0;

MobileGL/MG_Backend/DirectGLES/BackendObject_DirectGLES.cpp

@@ -18,6 +18,7 @@
 #include <MG_Util/Converters/MGToGL/TextureEnumConverter.h>
 #include <MG_Util/Converters/MGToStr/TextureEnumConverter.h>
 #include <MG_Util/Texture/TextureFormatProcessor.h>
+#include <MG_Util/Async/ShaderCompilePool.h>
 #include <Config.h>
 #include <algorithm>
 #include <cmath>
@@ -209,7 +210,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
                 reasons.push_back("GL_DEPTH_COMPONENT32 native probe failed on OpenGL ES");
             }
             if (options & PixelFormatNormalizeOptionBit::NoThreeChannelRenderTarget) {
-                reasons.push_back("no three-channel multisample storage format on OpenGL ES");
+                reasons.push_back("no colour-renderable three-channel format on OpenGL ES");
             }
             if (options & PixelFormatNormalizeOptionBit::NoSnorm16RenderTarget) {
                 reasons.push_back("EXT_render_snorm not supported");
@@ -552,26 +553,60 @@ namespace MobileGL::MG_Backend::DirectGLES {
 
                 for (SizeT targetIndex = 0; targetIndex < kFormatCapabilityTextureTargetCount; ++targetIndex) {
                     const auto target = static_cast<TextureTarget>(targetIndex);
-                    // A multisample texture can only ever be rendered into, so its storage format
+                    // Colour-attachable targets need a colour-renderable fallback; the ordinary
-                    // has to stay colour-renderable; the ordinary fallback for a three-channel
+                    // fallback for a three-channel format is another three-channel one, which ES
-                    // format is a three-channel one, which ES accepts as a texture but rejects as
+                    // accepts as a texture but never as an attachment. Recompute the fallback per
-                    // multisample storage. Recompute the fallback per target so those formats get
+                    // target so those formats get widened where the target demands it.
-                    // widened here and nowhere else.
+                    const Flags<PixelFormatNormalizeOptionBit> renderTargetOptions =
-                    Flags<PixelFormatNormalizeOptionBit> targetOptions;
+                        TextureImpl::GetRenderTargetNormalizeOptions(capabilities, targetIndex);
-                    if (IsGLESProbeMultisampleTarget(target)) {
+                    // Multisample storage has no three-channel form on ES at all, so its widening
-                        targetOptions |= PixelFormatNormalizeOptionBit::NoThreeChannelRenderTarget;
+                    // is unconditional and skips the native probe (which cannot succeed). Every
-                        if (!capabilities.SupportsRenderSnorm || !capabilities.SupportsNorm16Texture) {
+                    // other target keeps the widening on the DRIVER branch, behind the native
-                            targetOptions |= PixelFormatNormalizeOptionBit::NoSnorm16RenderTarget;
+                    // probe: `shouldProbeFallback = !nativeCreated || !nativeRenderable` below is
-                        }
+                    // what makes the substitution conditional on the driver actually refusing, so
-                    }
+                    // a driver that does render to a three-channel image keeps allocating it byte
+                    // for byte. That is a per-format runtime answer, NOT a desktop-vs-device
+                    // split: llvmpipe renders to GL_RGB16F but refuses GL_RGB8_SNORM, GL_SRGB8,
+                    // GL_RGB32F and the RGB integer formats, so the CI driver widens those eight
+                    // too. Re-run the retrace fixtures and the glcts suites on any change here.
+                    const Bool widenUnconditionally = IsGLESProbeMultisampleTarget(target);
                     GLESProbeFormatInfo fallbackInfo = outerFallbackInfo;
                     Bool hasForcedFallback = outerHasForcedFallback;
-                    if (targetOptions) {
+                    if (renderTargetOptions) {
-                        hasForcedFallback = BuildFallbackProbeFormatInfo(
+                        // Folded into the forced options only when a forced fallback already
-                            requestedInternalFormat, forcedOptions | targetOptions, true, fallbackInfo);
+                        // applies, so the render-target bits never *create* one: ANGLE's forced
-                        if (!hasForcedFallback) {
+                        // GL_RGB8_SNORM -> GL_RGB16F is still three-channel and still needs
-                            BuildFallbackProbeFormatInfo(requestedInternalFormat, driverOptions | targetOptions, false,
+                        // widening, but a non-ANGLE driver must not lose its native probe.
+                        const Flags<PixelFormatNormalizeOptionBit> forcedProbeOptions =
+                            (outerHasForcedFallback || widenUnconditionally) ? forcedOptions | renderTargetOptions
+                                                                             : forcedOptions;
+                        hasForcedFallback =
+                            BuildFallbackProbeFormatInfo(requestedInternalFormat, forcedProbeOptions, true,
                                                          fallbackInfo);
+                        if (!hasForcedFallback) {
+                            BuildFallbackProbeFormatInfo(requestedInternalFormat,
+                                                         driverOptions | renderTargetOptions, false, fallbackInfo);
+                        }
+                        // HONEST STATUS OF THE FORCED PATH. A forced fallback is only ever built
+                        // for ANGLE (GetForcedPixelFormatNormalizeOptions returns nothing for any
+                        // other renderer), and it SKIPS the native probe entirely - the widened
+                        // format is asserted rather than measured on this device. That assertion
+                        // is validated on exactly one configuration, the android-angle retrace
+                        // golden; it is NOT covered by the headless llvmpipe suites, which take
+                        // the driver branch below and prove nothing about ANGLE's answers. So log
+                        // the choice at INFO rather than the usual MGLOG_D caveat: on any other
+                        // ANGLE device the device report is the only evidence there is of which
+                        // storage format the image really got. Once per format on the ordinary 2D
+                        // target - repeating it for all ten targets would bury the report.
+                        if (hasForcedFallback && target == TextureTarget::Texture2D &&
+                            (MG_Util::TextureFormatProcessor::GetApplicablePixelFormatNormalizeOptions(
+                                 requestedInternalFormat, renderTargetOptions) &
+                             PixelFormatNormalizeOptionBit::NoThreeChannelRenderTarget)) {
+                            MGLOG_I("Three-channel widening (FORCED path, no native probe): %s stored as %s. "
+                                    "Reason: %s. Device-validated on the android-angle golden only.",
+                                    MG_Util::ConvertTextureInternalFormatToString(logicalFormat).c_str(),
+                                    ConvertFallbackInternalFormatToString(fallbackInfo.InternalFormat).c_str(),
+                                    fallbackInfo.Reason.c_str());
                         }
                     }
 
@@ -617,8 +652,26 @@ namespace MobileGL::MG_Backend::DirectGLES {
                 }
 
                 const SizeT renderbufferTargetIndex = GetRenderbufferFormatCapabilityTargetIndex();
-                Bool shouldProbeFallbackRenderbuffer = outerHasForcedFallback;
+                // A renderbuffer exists only to be attached, so it needs the same three-channel
-                if (!outerHasForcedFallback) {
+                // widening the colour-attachable texture targets get - and on the same terms: the
+                // native storage is probed first, so a driver that renders to it keeps it.
+                const Flags<PixelFormatNormalizeOptionBit> renderbufferOptions =
+                    TextureImpl::GetRenderTargetNormalizeOptions(capabilities, renderbufferTargetIndex);
+                GLESProbeFormatInfo renderbufferFallbackInfo = outerFallbackInfo;
+                Bool renderbufferHasForcedFallback = outerHasForcedFallback;
+                if (renderbufferOptions) {
+                    const Flags<PixelFormatNormalizeOptionBit> forcedProbeOptions =
+                        outerHasForcedFallback ? forcedOptions | renderbufferOptions : forcedOptions;
+                    renderbufferHasForcedFallback = BuildFallbackProbeFormatInfo(
+                        requestedInternalFormat, forcedProbeOptions, true, renderbufferFallbackInfo);
+                    if (!renderbufferHasForcedFallback) {
+                        BuildFallbackProbeFormatInfo(requestedInternalFormat, driverOptions | renderbufferOptions,
+                                                     false, renderbufferFallbackInfo);
+                    }
+                }
+
+                Bool shouldProbeFallbackRenderbuffer = renderbufferHasForcedFallback;
+                if (!renderbufferHasForcedFallback) {
                     const Bool nativeRenderbufferComplete =
                         ProbeRenderbuffer(gl, nativeInfo.InternalFormat, logicalFormat, false, 1);
                     if (nativeRenderbufferComplete) {
@@ -632,16 +685,16 @@ namespace MobileGL::MG_Backend::DirectGLES {
                         shouldProbeFallbackRenderbuffer = true;
                     }
                 }
-                if (shouldProbeFallbackRenderbuffer && outerFallbackInfo.InternalFormat != GL_UNKNOWN_MGL &&
+                if (shouldProbeFallbackRenderbuffer && renderbufferFallbackInfo.InternalFormat != GL_UNKNOWN_MGL &&
-                    ProbeRenderbuffer(gl, outerFallbackInfo.InternalFormat, logicalFormat, false, 1)) {
+                    ProbeRenderbuffer(gl, renderbufferFallbackInfo.InternalFormat, logicalFormat, false, 1)) {
                     if (AddCaveatFormatCaps(cache, renderbufferTargetIndex, formatIndex,
                                             GetRenderbufferFeatureCaps(logicalFormat))) {
-                        LogGLESFormatCaveat(logicalFormat, renderbufferTargetIndex, outerFallbackInfo);
+                        LogGLESFormatCaveat(logicalFormat, renderbufferTargetIndex, renderbufferFallbackInfo);
                     }
                     const Int maxSamples =
-                        GetGLESFormatMaxSamples(capabilities, logicalFormat, outerFallbackInfo.ImageFormat);
+                        GetGLESFormatMaxSamples(capabilities, logicalFormat, renderbufferFallbackInfo.ImageFormat);
-                    cache.SampleCounts[renderbufferTargetIndex][formatIndex] =
+                    cache.SampleCounts[renderbufferTargetIndex][formatIndex] = ProbeRenderbufferSampleCounts(
-                        ProbeRenderbufferSampleCounts(gl, outerFallbackInfo.InternalFormat, logicalFormat, maxSamples);
+                        gl, renderbufferFallbackInfo.InternalFormat, logicalFormat, maxSamples);
                 }
             }
         }
@@ -659,9 +712,9 @@ namespace MobileGL::MG_Backend::DirectGLES {
                     {
                         .TargetGLVersion = {4, 0, 0},   // GL target version
                         .TargetGLSLVersion = {4, 6, 0}, // Target Shading Language Version
-                        // Baseline advertisement (no timer queries / anisotropy yet); reconciled
+                        // Baseline advertisement (no runtime capabilities yet); reconciled once
-                        // once the ES capabilities exist, see UpdateAdvertisedCapabilityExtensions.
+                        // the ES capabilities exist, see UpdateAdvertisedCapabilityExtensions.
-                        .Extensions = BuildAdvertisedExtensions(false, false),
+                        .Extensions = BuildAdvertisedExtensions(false, false, false, false),
                         .IsCompatibilityProfile = false // Is Compatibility Profile
                     },
                 .StaticBackendCapability = {.AllowVSOnlyPrograms = false} // Backend Capability
@@ -681,9 +734,11 @@ namespace MobileGL::MG_Backend::DirectGLES {
         // thread can only observe the extension string after the
         // advertisement for its context has settled; rebuilding the whole
         // list keeps the re-run after a context recreation idempotent.
-        void UpdateAdvertisedCapabilityExtensions(Bool anisotropicFilteringSupported) {
+        void UpdateAdvertisedCapabilityExtensions(const MG_External::GLESCapabilities& capabilities) {
-            MutableRendererInfo().RendererGLInfo.Extensions =
+            MutableRendererInfo().RendererGLInfo.Extensions = BuildAdvertisedExtensions(
-                BuildAdvertisedExtensions(AreTimerQueriesSupported(), anisotropicFilteringSupported);
+                AreTimerQueriesSupported(), capabilities.SupportsTextureFilterAnisotropy,
+                capabilities.SupportsDrawIndirect,
+                capabilities.SupportsDrawIndirect && capabilities.SupportsBaseInstance);
         }
     } // namespace
 
@@ -726,11 +781,11 @@ namespace MobileGL::MG_Backend::DirectGLES {
             return false;
         }
         DirectGLES::SetGLESCapabilities(m_GLESCapabilities);
-        // Now that g_GLESCapabilities knows about GL_EXT_disjoint_timer_query and
+        // Now that g_GLESCapabilities knows the host extensions, entry points, and ES version,
-        // GL_EXT_texture_filter_anisotropic, reconcile the advertisement (see the comment on
+        // reconcile every runtime-gated advertisement (see the comment on
-        // UpdateAdvertisedCapabilityExtensions for why it cannot happen when the extension
+        // UpdateAdvertisedCapabilityExtensions for why this cannot happen when the list is first
-        // list is first built).
+        // built).
-        UpdateAdvertisedCapabilityExtensions(m_GLESCapabilities.SupportsTextureFilterAnisotropy);
+        UpdateAdvertisedCapabilityExtensions(m_GLESCapabilities);
         UpdateDynamicBackendParameters();
         PopulateFormatCapabilities(m_GLESFunctions, m_GLESCapabilities, MutableFormatCapabilities());
         PrintFormatCapabilities(GetFormatCapabilities());
@@ -871,11 +926,13 @@ namespace MobileGL::MG_Backend::DirectGLES {
         return MutableRendererInfo();
     }
 
-    Vector<GLExtension> BuildAdvertisedExtensions(Bool timerQueriesSupported, Bool anisotropicFilteringSupported) {
+    Vector<GLExtension> BuildAdvertisedExtensions(Bool timerQueriesSupported, Bool anisotropicFilteringSupported,
+                                                  Bool drawIndirectSupported,
+                                                  Bool nonZeroIndirectBaseInstanceSupported) {
         Vector<GLExtension> extensions = {
             V_OpenGL30, V_OpenGL31, V_OpenGL32, V_OpenGL33, V_OpenGL40, E_GL_ARB_draw_buffers_blend,
             E_GL_ARB_compute_shader, E_GL_ARB_shader_storage_buffer_object, E_GL_ARB_shader_image_load_store,
-            E_GL_ARB_program_interface_query, E_GL_ARB_framebuffer_object, E_GL_EXT_framebuffer_object,
+            E_GL_ARB_clear_buffer_object, E_GL_ARB_program_interface_query, E_GL_ARB_framebuffer_object, E_GL_EXT_framebuffer_object,
             E_GL_ARB_depth_texture, E_GL_ARB_buffer_storage, E_GL_ARB_texture_storage,
             E_GL_ARB_texture_storage_multisample, E_GL_ARB_clear_texture, E_GL_ARB_direct_state_access,
             E_GL_ARB_multi_draw_indirect, E_GL_ARB_indirect_parameters, E_GL_ARB_shader_draw_parameters,
@@ -887,10 +944,59 @@ namespace MobileGL::MG_Backend::DirectGLES {
             // picks a whole different shader for draw_buffers without
             // explicit_attrib_location. DirectVulkan advertises both.
             E_GL_ARB_explicit_attrib_location, E_GL_ARB_texture_multisample, E_GL_ARB_shader_image_size,
+            // Core since GL 3.1 and implemented for every version advertised here. The string
+            // matters because applications gate the ENTRY POINTS on it rather than on the
+            // version: a caller that finds the extension missing never resolves
+            // glGetUniformBlockIndex / glUniformBlockBinding, and one that then uses uniform
+            // blocks anyway calls through a null pointer.
+            E_GL_ARB_uniform_buffer_object,
+            // Sampling the stencil aspect through DEPTH_STENCIL_TEXTURE_MODE. Core from 4.3,
+            // so on a 4.0 context the string is the only way to reach it. The host ES driver
+            // has had the same texture parameter since ES 3.1, which every device MobileGL
+            // runs on provides.
+            E_GL_ARB_stencil_texturing,
             // Advertised with GL_NUM_PROGRAM_BINARY_FORMATS = 0, which the
             // extension explicitly permits. It is also the only thing that
             // exposes glProgramParameteri before GL 4.1.
             E_GL_ARB_get_program_binary};
+        // Minecraft 26.3 checks this prerequisite before it even considers
+        // GL_ARB_multi_draw_indirect. ES 3.1 supplies both single-draw entry points; the loader
+        // folds the version and pointer checks into SupportsDrawIndirect.
+        if (drawIndirectSupported) {
+            extensions.push_back(E_GL_ARB_draw_indirect);
+        }
+        // ARB_base_instance also defines the last word of an indirect command. Direct calls are
+        // emulated on every Espryt device, but without host GL_EXT_base_instance a native indirect
+        // draw cannot shift divisor attributes by a GPU-authored non-zero value, so do not promise
+        // that incomplete case.
+        if (drawIndirectSupported && nonZeroIndirectBaseInstanceSupported) {
+            extensions.push_back(E_GL_ARB_base_instance);
+        }
+        // GL_KHR_parallel_shader_compile is MobileGL's own capability, not the host ES
+        // driver's: the compiler threads are MobileGL's, and glCompileShader/glLinkProgram
+        // are serviced entirely inside the frontend. Whether the device driver advertises
+        // the string is irrelevant here (the POST reports it separately, for the day the
+        // driver-side link is what gets parallelised).
+        //
+        // Gated on the async flag deliberately, and this is the whole reason the gate
+        // exists. Advertising the string is the one part of asynchronous compilation that a
+        // recorded trace can never cover: Iris and Sodium change their SUBMISSION SCHEDULE
+        // the moment they see it - they enqueue whole pipeline batches and poll
+        // GL_COMPLETION_STATUS_KHR instead of compiling one program at a time - so
+        // MOBILEGL_ASYNC_SHADER_COMPILE=0 has to withdraw the application-visible behaviour
+        // change as well as the threading, or the kill switch would only be half a switch.
+        if (MG_Util::Async::AsyncShaderCompileEnabled()) {
+            extensions.push_back(E_GL_KHR_parallel_shader_compile);
+        }
+        // GL_ARB_gpu_shader_fp64 is opt-in (MOBILEGL_ADVERTISE_FP64). Every `double` in a
+        // shader compiles and runs already - it is narrowed to 32 bits before the module
+        // reaches this backend - so an application that simply uses doubles needs nothing
+        // advertised. What the extension additionally promises is 64-bit PRECISION, which no
+        // mobile GPU has and the narrowing cannot fake, so advertising it by default would
+        // make an application that checks the string take a path MobileGL cannot honour.
+        if (MG_Config::Features.AdvertiseFp64) {
+            extensions.push_back(E_GL_ARB_gpu_shader_fp64);
+        }
         // Only advertised when the device driver actually has usable timer queries
         // (GL_EXT_disjoint_timer_query plus its entry points) and the
         // MOBILEGL_DISABLE_TIMERQUERY escape hatch is off.
@@ -933,6 +1039,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
             funcsTable.GL.MultiDrawElementsIndirect = MultiDrawElementsIndirect;
             funcsTable.GL.MultiDrawElementsIndirectCount = MultiDrawElementsIndirectCount;
             funcsTable.GL.MultiDrawArraysIndirect = MultiDrawArraysIndirect;
+            funcsTable.GL.MultiDrawArraysIndirectCount = MultiDrawArraysIndirectCount;
             funcsTable.GL.DrawRangeElementsBaseVertex = DrawRangeElementsBaseVertex;
             funcsTable.GL.DrawRangeElements = DrawRangeElements;
             funcsTable.GL.DrawElementsInstancedBaseVertexBaseInstance = DrawElementsInstancedBaseVertexBaseInstance;
@@ -951,12 +1058,6 @@ namespace MobileGL::MG_Backend::DirectGLES {
             funcsTable.GL.GetIntegeri_v = GetIntegeri_v;
             funcsTable.GL.GetInteger64i_v = GetInteger64i_v;
             funcsTable.GL.GetProgramiv = GetProgramiv;
-            funcsTable.GL.GetProgramInterfaceiv = GetProgramInterfaceiv;
-            funcsTable.GL.GetProgramResourceIndex = GetProgramResourceIndex;
-            funcsTable.GL.GetProgramResourceName = GetProgramResourceName;
-            funcsTable.GL.GetProgramResourceiv = GetProgramResourceiv;
-            funcsTable.GL.GetProgramResourceLocation = GetProgramResourceLocation;
-            funcsTable.GL.GetProgramResourceLocationIndex = GetProgramResourceLocationIndex;
             funcsTable.GL.ShaderStorageBlockBinding = ShaderStorageBlockBinding;
             funcsTable.GL.Clear = Clear;
             funcsTable.GL.ClearBufferfi = ClearBufferfi;
@@ -1088,6 +1189,16 @@ namespace MobileGL::MG_Backend::DirectGLES {
         m_dynamicParameters.MaxComputeUniformBlocks = m_GLESCapabilities.MaxComputeUniformBlocks;
         m_dynamicParameters.MaxComputeWorkGroupInvocations = m_GLESCapabilities.MaxComputeWorkGroupInvocations;
         m_dynamicParameters.MaxShaderStorageBufferBindings = m_GLESCapabilities.MaxShaderStorageBufferBindings;
+        // This is the number glGetIntegerv(GL_MAX_TEXTURE_BUFFER_SIZE) hands the application, and
+        // on a host without buffer textures it is knowingly a floor MobileGL cannot honour rather
+        // than a driver answer (m_GLESCapabilities.MaxTextureBufferSizeIsDriverReported says
+        // which). Reporting 0 instead was considered and rejected: MobileGL advertises an OpenGL
+        // 4.x context, where buffer textures are core and the limit has a spec minimum of 65536,
+        // so 0 is not a legal answer and applications are not written to survive it. GL offers no
+        // way to say "this core feature is missing", so the honesty is carried outside the limit:
+        // FillInGLESCapabilities logs the tier, glTexBuffer and the program build each name the
+        // missing capability at MGLOG_I, and the driver POST carries a "Buffer textures" row that
+        // FAILs on this tier.
         m_dynamicParameters.MaxTextureBufferSize = m_GLESCapabilities.MaxTextureBufferSize;
         m_dynamicParameters.TextureBufferOffsetAlignment = m_GLESCapabilities.TextureBufferOffsetAlignment;
         m_dynamicParameters.MaxUniformBufferBindings = m_GLESCapabilities.MaxUniformBufferBindings;
@@ -1112,8 +1223,28 @@ namespace MobileGL::MG_Backend::DirectGLES {
         // SyncAttachmentObject routes a layered upload target to glFramebufferTextureLayer with the
         // attachment's layer passed through, so this backend really does render to the layer it was
         // given - provided the driver resolved the entry point at all.
-        m_dynamicParameters.SupportsPerLayerFramebufferAttachment =
+        // SyncAttachmentObject (Managers.cpp, the glFramebufferTextureLayer branch) routes exactly
-            DirectGLES::g_GLESFuncs.glFramebufferTextureLayer != nullptr;
+        // five upload targets to glFramebufferTextureLayer with the attachment's layer passed
+        // through, so this backend really does render to the layer it was given - provided the driver
+        // resolved the entry point at all. The cube map array is the one target that also needs
+        // ES-level support before it has any storage to attach.
+        m_dynamicParameters.PerLayerFramebufferAttachmentTargets = 0;
+        if (DirectGLES::g_GLESFuncs.glFramebufferTextureLayer != nullptr) {
+            using DynParams = MG_Backend::DynamicBackendParameters;
+            m_dynamicParameters.PerLayerFramebufferAttachmentTargets |=
+                DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::Texture3D) |
+                DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::Texture1DArray) |
+                DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::Texture2DArray) |
+                DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::Texture2DMultisampleArray);
+            if (m_GLESCapabilities.SupportsTextureCubeMapArray) {
+                m_dynamicParameters.PerLayerFramebufferAttachmentTargets |=
+                    DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::TextureCubeMapArray);
+            }
+        }
+        // Not a driver question and never will be: OpenGL ES has no double-precision vertex format
+        // and ESSL has no fp64 type to consume one with, so a 64-bit vertex attribute has nowhere to
+        // land on this backend regardless of what the driver underneath happens to support.
+        m_dynamicParameters.SupportsFloat64VertexAttributes = false;
         m_dynamicParameters.MaxDrawBuffers = m_GLESCapabilities.MaxDrawBuffers;
         m_dynamicParameters.MaxColorAttachments = m_GLESCapabilities.MaxColorAttachments;
         m_dynamicParameters.MaxClipDistances = m_GLESCapabilities.MaxClipDistances;

MobileGL/MG_Backend/DirectGLES/BackendObject_DirectGLES.h

@@ -67,9 +67,12 @@ namespace MobileGL::MG_Backend::DirectGLES {
     const RendererInfo& GetRendererIdentity();
 
     // The full OpenGL extension list Espryt advertises (glGetString(GL_EXTENSIONS))
-    // for a device whose timer queries / anisotropic filtering are (or are not) usable.
+    // for a device whose timer queries / anisotropic filtering / native indirect draws /
+    // non-zero indirect baseInstance semantics are (or are not) usable.
     // The MOBILEGL_DISABLE_TIMERQUERY escape hatch is applied inside.
-    Vector<GLExtension> BuildAdvertisedExtensions(Bool timerQueriesSupported, Bool anisotropicFilteringSupported);
+    Vector<GLExtension> BuildAdvertisedExtensions(Bool timerQueriesSupported, Bool anisotropicFilteringSupported,
+                                                  Bool drawIndirectSupported,
+                                                  Bool nonZeroIndirectBaseInstanceSupported);
 
     // Format: <OpenGL ES Renderer>, OpenGL ES <Major>.<Minor> — the exact string an
     // initialized backend returns from GetBackendAPIVersionString (and that ends up

MobileGL/MG_Backend/DirectGLES/DirectGLES.h

@@ -40,6 +40,8 @@ namespace MobileGL::MG_Backend::DirectGLES {
     void MultiDrawElementsIndirectCount(GLenum mode, GLenum type, const void* indirect, GLintptr drawcount,
                                         GLsizei maxdrawcount, GLsizei stride);
     void MultiDrawArraysIndirect(GLenum mode, const void* indirect, GLsizei drawcount, GLsizei stride);
+    void MultiDrawArraysIndirectCount(GLenum mode, const void* indirect, GLintptr drawcount, GLsizei maxdrawcount,
+                                      GLsizei stride);
     void DrawRangeElementsBaseVertex(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type,
                                      const void* indices, GLint basevertex);
     void DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type, const void* indices);
@@ -92,15 +94,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
     void GetIntegeri_v(GLenum target, GLuint index, GLint* data);
     void GetInteger64i_v(GLenum target, GLuint index, GLint64* data);
     void GetProgramiv(GLuint program, GLenum pname, GLint* params);
-    void GetProgramInterfaceiv(GLuint program, GLenum programInterface, GLenum pname, GLint* params);
+    void ShaderStorageBlockBinding(GLuint program, const GLchar* storageBlockName, GLuint storageBlockBinding);
-    GLuint GetProgramResourceIndex(GLuint program, GLenum programInterface, const GLchar* name);
-    void GetProgramResourceName(GLuint program, GLenum programInterface, GLuint index, GLsizei bufSize, GLsizei* length,
-                                GLchar* name);
-    void GetProgramResourceiv(GLuint program, GLenum programInterface, GLuint index, GLsizei propCount,
-                              const GLenum* props, GLsizei bufSize, GLsizei* length, GLint* params);
-    GLint GetProgramResourceLocation(GLuint program, GLenum programInterface, const GLchar* name);
-    GLint GetProgramResourceLocationIndex(GLuint program, GLenum programInterface, const GLchar* name);
-    void ShaderStorageBlockBinding(GLuint program, GLuint storageBlockIndex, GLuint storageBlockBinding);
     Bool InitWindowSurface(NativeWindowType window);
     Bool InitPbufferSurface(EGLint width, EGLint height);
     Bool MakeCurrent();
@@ -125,6 +119,24 @@ namespace MobileGL::MG_Backend::DirectGLES {
     // capability read needs no current ES context, and it stays false until
     // the ES capabilities have been filled in.
     Bool AreTimerQueriesSupported();
+    // True when the host ES driver can back a GL_TEXTURE_BUFFER at all - ES 3.2 core, or
+    // EXT/OES_texture_buffer, with glTexBuffer resolved. Desktop GL has had buffer textures as
+    // core since 3.1, so the frontend advertises them unconditionally and an app may call
+    // glTexBuffer whenever it likes; this is the only thing standing between that call and a
+    // null entry point. False also means every shader declaring a samplerBuffer is
+    // uncompilable on this driver, which the program build reports by name.
+    Bool AreBufferTexturesSupported();
+    // Human-readable name of the buffer-texture tier for diagnostics and the driver POST:
+    // "core (ES 3.2)", "GL_EXT_texture_buffer", "GL_OES_texture_buffer" or "unsupported".
+    const char* GetBufferTextureTierName();
+    // glTexBuffer / glTexBufferRange through whichever spelling this driver's buffer-texture
+    // support actually ships: the unsuffixed names are ES 3.2 core, while an EXT/OES driver
+    // exports glTexBuffer{,Range}EXT / OES. Callers must have checked
+    // AreBufferTexturesSupported() first. CallTexBufferRange reports whether it could honour
+    // the range - no tier is required to expose the range form, and the whole-buffer form is
+    // the documented fallback.
+    void CallTexBuffer(GLenum target, GLenum internalFormat, GLuint buffer);
+    Bool CallTexBufferRange(GLenum target, GLenum internalFormat, GLuint buffer, GLintptr offset, GLsizeiptr size);
     // GL timer-query objects, backed by GL_EXT_disjoint_timer_query. The
     // creators return null (the frontend then falls back to an immediately
     // available zero result) when the calling thread does not own the ES
@@ -160,6 +172,11 @@ namespace MobileGL::MG_Backend::DirectGLES {
     // A buffer retired during frame N is safe to recycle once CompletedFrameSerial() >= N.
     Uint64 CurrentFrameSerial();
     Uint64 CompletedFrameSerial();
+    // Block (up to timeoutNs) until the given frame serial provably retired on the
+    // GPU, using the per-frame fence ring. False when no usable fence covers the
+    // serial (fence-less context, foreign thread, or the slot was recycled);
+    // completion state is untouched in that case.
+    Bool WaitForFrameSerialCompleted(Uint64 serial, Uint64 timeoutNs);
     // Applies (or defers until the window surface exists) the app-requested
     // eglSwapInterval on the native EGL surface.
     void SetSwapInterval(Int interval);
@@ -177,6 +194,11 @@ namespace MobileGL::MG_Backend::DirectGLES {
 
     namespace XfbImpl {
         Bool AreTransformFeedbacksSupported();
+        // True while a capture span is open on the current transform feedback object
+        // (frontend Begin seen and not paused), whether or not the deferred driver-side
+        // Begin has been issued yet. Draw paths that would restructure the primitive
+        // stream, or that need to dispatch compute mid-draw, decline while it is set.
+        Bool IsCaptureSpanOpen();
         void BeginTransformFeedback(GLenum primitiveMode);
         void EndTransformFeedback();
         void PauseTransformFeedback();
@@ -186,6 +208,26 @@ namespace MobileGL::MG_Backend::DirectGLES {
         void OnBackendContextDestroyed();
     } // namespace XfbImpl
 
+    namespace RenderStateImpl {
+        // Pushes the frontend's render-state block to the ES driver, diffed against what was
+        // last pushed.
+        //
+        // `forColorClear` names the CALLER, and the only thing it changes is the colour write
+        // mask handed to the driver. A draw into a colour attachment the backend widened from
+        // three channels to four gets that buffer's alpha channel masked OFF, so nothing can
+        // move the stored alpha away from the 1.0 the application's three-channel format
+        // implies (see FramebufferImpl::g_alphaWidenedDrawBufferMask). A CLEAR is how that 1.0
+        // gets there in the first place, so it must be allowed to write alpha - hence the flag
+        // rather than an unconditional doctoring. It is part of the sync memo, so a clear
+        // followed by a draw re-pushes the mask instead of early-outing on an unchanged
+        // frontend version.
+        //
+        // The application's own colour mask is never modified: glGet(GL_COLOR_WRITEMASK)
+        // answers from the frontend state, which this function only reads.
+        void SyncRenderState(Bool forColorClear = false);
+        void InvalidateSyncedRenderState();
+    } // namespace RenderStateImpl
+
     extern MG_External::EGLFunctionsTable g_EGLFuncs;
     extern MG_External::GLESFunctionsTable g_GLESFuncs;
     extern MG_External::GLESCapabilities g_GLESCapabilities;

MobileGL/MG_Backend/DirectGLES/Managers.h

@@ -21,45 +21,151 @@ namespace MobileGL::MG_Backend::DirectGLES {
     String EmulateBaseInstanceInVertexShader(String source, GLenum shaderType);
     String PromoteDrawParameterGlobalsToUniforms(String source, GLenum shaderType);
 
+    // True once the process has entered exit(): past that point the EGL library and
+    // the driver may already be unloaded, so a backend twin's destructor must not
+    // call into g_GLESFuncs (the observed crash is a jump through an unmapped driver
+    // pointer from __run_exit_handlers) nor touch statics in other TUs (cross-TU
+    // destruction order is unspecified). Deliberate leak: the process is exiting and
+    // the driver reclaims GPU objects. The flag is set by a std::atexit handler that
+    // EnsureProcessTeardownSentinel() registers lazily on first registry use - by
+    // then every static everywhere has finished constructing, so this handler is
+    // guaranteed to run BEFORE any static destructor (atexit is LIFO). A destructor
+    // hook on the registry itself was tried first and is WRONG: tests and cache
+    // resets destroy temporary registry instances mid-run, which would latch the
+    // flag while the process is very much alive.
+    Bool InProcessTeardown();
+    void EnsureProcessTeardownSentinel();
+
+    // Generation of the backend ES context that owns the driver ids currently handed
+    // out. Bumped exactly once per DestroyEGLContext. Every backend twin that owns a
+    // driver name (texture, framebuffer, renderbuffer, sampler) stamps this at
+    // construction and compares it in its destructor: a twin outliving its context
+    // must NOT glDelete* its id, because a successor context may already have recycled
+    // that name and the delete would take out a live object of the new context.
+    extern Uint g_backendContextGeneration;
+
+    // Which optional pieces of state a draw needs synchronized before it is issued.
+    // Index/indirect buffer syncs and the instancing-related work are skipped for
+    // draws that provably cannot read them.
+    enum class DrawSyncBit : Uint32 {
+        None = 0,
+        IndexBuffer = 1 << 0,
+        IndirectBuffer = 1 << 1,
+        Instancing = 1 << 2
+    };
+    // Deliberately the shared Flags<> rather than hand-written operators for this enum:
+    // a namespace-local operator| here would hide MobileGL::operator|(Bit, Bit) from
+    // every other scoped-enum flag set used inside this namespace.
+    using DrawSyncFlags = Flags<DrawSyncBit>;
+
+    // The GL-defined indirect command layouts, byte-identical to what the driver reads
+    // out of a GL_DRAW_INDIRECT_BUFFER. Also the staging layout the multi-draw emulation
+    // synthesizes commands into.
+    struct DrawElementsIndirectCommand {
+        Uint32 count = 0;
+        Uint32 instanceCount = 0;
+        Uint32 firstIndex = 0;
+        Int32 baseVertex = 0;
+        Uint32 baseInstance = 0;
+    };
+
+    struct DrawArraysIndirectCommand {
+        Uint32 count = 0;
+        Uint32 instanceCount = 0;
+        Uint32 first = 0;
+        Uint32 baseInstance = 0;
+    };
+
+    // Brings the whole draw-relevant frontend state onto the native ES context and binds
+    // the program; every GL draw entry point calls it exactly once before issuing draws.
+    void PrepareForDraw(DrawSyncFlags syncBits);
+    // GLES core supports only GL_PRIMITIVE_RESTART_FIXED_INDEX. Throws when the app enabled
+    // the arbitrary GL_PRIMITIVE_RESTART with a non-fixed index for this index type.
+    void CheckPrimitiveRestartSupported(GLenum indexType);
+    // Feed the current program's gl_BaseInstance / gl_DrawID / gl_BaseVertex emulation
+    // uniforms. All are no-ops when the program does not read the corresponding builtin.
+    void SetCurrentBaseInstance(Uint32 baseInstance);
+    void SetCurrentDrawID(Uint32 drawId);
+    // GL's gl_BaseVertex is the base-vertex parameter of an indexed draw and zero for every
+    // command that has none - including all the DrawArrays forms - so every draw path that
+    // does not carry one must leave this at zero rather than inherit the last draw's value.
+    void SetCurrentBaseVertex(Int32 baseVertex);
+    // True when the current program actually reads gl_DrawID, i.e. when a batched
+    // (single driver call) multi-draw tier would have to feed it one value for the whole
+    // batch and would therefore be wrong.
+    Bool CurrentProgramReadsDrawID();
+    // Same question for gl_BaseVertex: a batched multi-draw tier cannot give each sub-draw
+    // its own base vertex through a uniform either.
+    Bool CurrentProgramReadsBaseVertex();
+    // Both of the above, conservatively, for a caller that must decide BEFORE PrepareForDraw
+    // has synced the program - where "does not read it" is indistinguishable from "cannot be
+    // asked yet". Answers true whenever the backend twin is missing or predates the current
+    // link.
+    Bool CurrentProgramMayNeedPerSubDrawBuiltins(Bool batchCarriesBaseVertices);
+
     template <typename StateObject, typename BackendObject>
     class StateBackendObjectRegistry {
     public:
+
         using StatePtr = SharedPtr<StateObject>;
         using StateWeakPtr = std::weak_ptr<StateObject>;
         using BackendPtr = SharedPtr<BackendObject>;
-        using BackendMap = UnorderedMap<StateObject*, BackendPtr>;
+
-        using StateRefMap = UnorderedMap<StateObject*, StateWeakPtr>;
+        // The backend twin and the weak reference that decides whether the raw key still
+        // names the state object the twin was built for. Both live in one entry: a
+        // separate liveness map answered nothing the backend probe had not already found
+        // and cost a second hash lookup on every Find, which the draw path runs ~10 times.
+        struct Entry {
+            BackendPtr backend;
+            StateWeakPtr stateRef;
+        };
+        using BackendMap = UnorderedMap<StateObject*, Entry>;
         using iterator = typename BackendMap::iterator;
         using const_iterator = typename BackendMap::const_iterator;
 
         BackendPtr& GetOrCreate(const StatePtr& stateObj) {
             MOBILEGL_ASSERT(stateObj != nullptr, "State object must not be null");
 
-            auto* key = stateObj.get();
+            // Twin creation is the moment a driver-owned id starts needing a guarded
-            auto trackedStateIt = m_stateRefs.find(key);
+            // destructor; cold path, so the once-guard costs nothing per draw.
-            if (trackedStateIt != m_stateRefs.end() && trackedStateIt->second.expired()) {
+            EnsureProcessTeardownSentinel();
-                EraseByKey(key);
+            auto& entry = m_entries[stateObj.get()];
+            if (entry.stateRef.expired()) {
+                // The previous owner of this address is gone and the allocator handed it
+                // to a new object: its twin describes ids the new state object never made.
+                entry.backend.reset();
             }
-            m_stateRefs[key] = stateObj;
+            entry.stateRef = stateObj;
-            return m_backendObjects[key];
+            return entry.backend;
         }
 
-        iterator find(StateObject* stateObj) {
+        // Null when no live state object owns this key. The result points into the map, so
-            if (!IsAlive(stateObj)) {
+        // it stays valid only until the next GetOrCreate/Find/CollectGarbage on this registry.
-                EraseByKey(stateObj);
+        // Take that literally, including for Find: the map is open-addressed and erases by
-                return m_backendObjects.end();
+        // shifting the rest of the probe cluster into the hole, so an erase relocates entries
+        // OTHER than the erased one - and Find erases, whenever it lands on a key whose state
+        // object has expired. Callers that need the twin across another registry call must copy
+        // the BackendPtr out (or keep only the pointee, which is heap-allocated and never moves).
+        BackendPtr* Find(StateObject* stateObj) {
+            const auto entryIt = m_entries.find(stateObj);
+            if (entryIt == m_entries.end()) {
+                return nullptr;
             }
-            return m_backendObjects.find(stateObj);
+            if (entryIt->second.stateRef.expired()) {
+                m_entries.erase(entryIt);
+                return nullptr;
+            }
+            return &entryIt->second.backend;
         }
 
-        const_iterator find(StateObject* stateObj) const {
+        const BackendPtr* Find(StateObject* stateObj) const {
-            return const_cast<StateBackendObjectRegistry*>(this)->find(stateObj);
+            return const_cast<StateBackendObjectRegistry*>(this)->Find(stateObj);
         }
 
-        iterator begin() { return m_backendObjects.begin(); }
+        iterator begin() { return m_entries.begin(); }
-        const_iterator begin() const { return m_backendObjects.begin(); }
+        const_iterator begin() const { return m_entries.begin(); }
-        iterator end() { return m_backendObjects.end(); }
+        iterator end() { return m_entries.end(); }
-        const_iterator end() const { return m_backendObjects.end(); }
+        const_iterator end() const { return m_entries.end(); }
 
         void CollectGarbageIfNeeded() {
             ++m_gcTick;
@@ -73,19 +179,6 @@ namespace MobileGL::MG_Backend::DirectGLES {
         void CollectGarbageNow() { CollectGarbage(); }
 
     private:
-        bool IsAlive(StateObject* stateObj) const {
-            const auto trackedStateIt = m_stateRefs.find(stateObj);
-            if (trackedStateIt == m_stateRefs.end()) {
-                return false;
-            }
-            return !trackedStateIt->second.expired();
-        }
-
-        void EraseByKey(StateObject* stateObj) {
-            m_stateRefs.erase(stateObj);
-            m_backendObjects.erase(stateObj);
-        }
-
         void CollectGarbage() {
             if (m_isCollecting) {
                 return;
@@ -94,16 +187,15 @@ namespace MobileGL::MG_Backend::DirectGLES {
             m_isCollecting = true;
 
             Vector<StateObject*> staleKeys;
-            staleKeys.reserve(m_stateRefs.size());
+            staleKeys.reserve(m_entries.size());
-            for (const auto& [stateKey, stateWeakRef] : m_stateRefs) {
+            for (const auto& [stateKey, entry] : m_entries) {
-                if (stateWeakRef.expired()) {
+                if (entry.stateRef.expired()) {
                     staleKeys.push_back(stateKey);
                 }
             }
 
             for (auto* stateKey : staleKeys) {
-                m_stateRefs.erase(stateKey);
+                m_entries.erase(stateKey);
-                m_backendObjects.erase(stateKey);
             }
 
             m_isCollecting = false;
@@ -111,8 +203,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
 
     private:
         static constexpr Uint32 kGCInterval = 1024;
-        StateRefMap m_stateRefs;
+        BackendMap m_entries;
-        BackendMap m_backendObjects;
         Uint32 m_gcTick = 0;
         Bool m_isCollecting = false;
     };
@@ -120,6 +211,43 @@ namespace MobileGL::MG_Backend::DirectGLES {
     namespace BufferImpl {
         const GLenum TempBufferTarget = GL_ARRAY_BUFFER;
 
+        // --- Buffer-mutation epoch -------------------------------------------------
+        // Manager-wide monotonic counter: it moves whenever ANY buffer resource may
+        // have gone from draw-clean to dirty. Draw-path memos read it once per pass
+        // (CurrentBufferMutationEpoch, acquire), re-run their IsBufferDrawClean
+        // probes only when it moved, and stamp the PRE-pass value after a pass in
+        // which every probe came up clean - so a concurrent bump lands strictly
+        // after the stamped value and forces a re-probe on the next pass no matter
+        // how the probe interleaved with the mutation. Conservative-correct: a bump
+        // never skips work, it only re-runs the probes once.
+        //
+        // Every clean->dirty transition path bumps it (BumpBufferMutationEpoch,
+        // release, AFTER the mutation lands so an acquire reader that still sees
+        // the old epoch cannot have missed the mutation):
+        //   * the frontend BufferBackendOps table - Respecify, SubData,
+        //     FlushMappedRange, AcquirePersistentMap, ReadbackFromGpu, OnDestroy -
+        //     which every frontend change-serial bump and every pending-range
+        //     queueing reaches while ops are registered (upload, orphan/respecify,
+        //     map flush/unmap writeback, persistent-map adoption, delete/pooling);
+        //   * backend-initiated shadow writebacks that bump the frontend change
+        //     serial without an op: transform-feedback capture readback
+        //     (XfbImpl::ReadbackCapturedRanges and the scatter path) and every
+        //     pack-PBO WritebackFromBackend site (glReadPixels/glGetTexImage);
+        //   * RegisterBufferBackendOps/UnregisterBufferBackendOps - while ops are
+        //     unregistered, frontend writes advance serials silently, so both edges
+        //     of that window re-open every memo;
+        //   * OnBackendContextDestroyed - the buffer context generation moved, so
+        //     every previously clean resource is invalid.
+        // NOT bumped (cleanliness provably unchanged): MarkGpuWritten (the backend
+        // copy is authoritative; IsBufferDrawClean does not consult it),
+        // NotifyContentWrite on a GPU-resident buffer (persistent-mapped resources
+        // are clean by construction), and EnsureBufferResource itself (it only
+        // repairs toward clean). A non-persistent map (draws on it are GL errors
+        // the frontend rejects) sets IsMapped without an op; persistent maps reach
+        // AcquirePersistentMap or (FLUSH_EXPLICIT) publish only via FlushMappedRange.
+        Uint64 CurrentBufferMutationEpoch();
+        void BumpBufferMutationEpoch();
+
         // The DirectGLES storage behind one frontend buffer. Owned (refcounted) by
         // the frontend BufferObject; immediate BufferBackendOps keep it current, so
         // draw-time "sync" reduces to ensuring the storage exists.
@@ -145,6 +273,12 @@ namespace MobileGL::MG_Backend::DirectGLES {
             Bool pendingRespecify = false;
             VecRange1D pendingRanges;
             std::mutex pendingMutex;
+            // Buffer-mutation epoch (see CurrentBufferMutationEpoch) at which this
+            // resource last probed IsBufferDrawClean == true, 0 = never (epochs start
+            // at 1). Written only on the draw thread; per-draw resource consumers
+            // (the UBO binding walk) skip the probe while their pre-pass epoch read
+            // matches, exactly like the per-VAO memo stamps.
+            Uint64 drawCleanEpoch = 0;
             // Zero-copy coherent persistent map (EXT_buffer_storage): the GL store is
             // immutable, persistently+coherently mapped, and persistentPtr is what the app
             // (and the frontend PipeResource) write into directly. While set, draw-time
@@ -152,6 +286,14 @@ namespace MobileGL::MG_Backend::DirectGLES {
             // context loss.
             Bool persistentMapped = false;
             void* persistentPtr = nullptr;
+            // The GL store behind `id` was created with glBufferStorageEXT and is
+            // therefore IMMUTABLE - glBufferData cannot respecify it and it must never be
+            // recycled through the size-keyed buffer pool. Tracked separately from
+            // persistentMapped because the two come apart: a glMapBufferRange that fails
+            // after its glBufferStorageEXT succeeded leaves immutable storage behind with
+            // no map, and a respecification then has to retire the id rather than hand it
+            // to glBufferData, which the driver would silently refuse.
+            Bool immutableStorage = false;
         };
 
         // Registered as the frontend's BufferBackendOps at backend init and on
@@ -170,6 +312,17 @@ namespace MobileGL::MG_Backend::DirectGLES {
         GLESBufferResource* EnsureBufferResource(const SharedPtr<MG_State::GLState::BufferObject>& bufferObject);
         // Existing resource or nullptr; performs no GL calls.
         GLESBufferResource* GetBufferResource(MG_State::GLState::BufferObject* bufferObject);
+        // True when EnsureBufferResource(frontend) would provably fall straight through
+        // every branch and do no work — i.e. `resource` is still the frontend's own
+        // resource, its id belongs to the live ES context, and either it is the
+        // zero-copy coherent persistent store (draw-time sync is a no-op by design) or
+        // the storage is initialized at the right size with no pending ops and a synced
+        // change serial while the buffer is not mapped (an active map may owe a
+        // per-draw persistent-range push, so it always takes the full path).
+        // `frontend` must be non-null and alive; the caller guarantees that by holding
+        // (or shadowing something that holds) a SharedPtr to it. Enables the per-VAO
+        // resolved-buffers memo to skip EnsureBufferResource on clean static buffers.
+        Bool IsBufferDrawClean(const MG_State::GLState::BufferObject* frontend, const GLESBufferResource* resource);
 
         // Deletes GL buffers whose owning frontend objects died (possibly on a
         // thread without a current ES context). Called from draw-time sync.
@@ -193,6 +346,14 @@ namespace MobileGL::MG_Backend::DirectGLES {
         // client-attribute staging buffers): scrub every buffer-binding shadow that
         // could false-skip when the name is recycled.
         void NoteBufferIdDeleted(Uint id);
+        // Bumped whenever a live GLESBufferResource's driver id is retired and re-minted
+        // while its frontend buffer stays alive (persistent-map adoption, immutable-store
+        // retire). The VAO twins' baked glVertexAttribPointer / element-array bindings
+        // key on FRONTEND versions, which a backend-side re-mint does not move - without
+        // this generation the driver VAO would keep fetching through the deleted id (or
+        // its retained store) forever. Compared and stamped by
+        // BackendVertexArrayObject::SyncToBackend.
+        extern Uint64 g_bufferBackendIdGeneration;
         // Redundant-bind cache for INDEXED buffer bindings (glBindBufferBase/Range on
         // GL_UNIFORM_BUFFER / GL_SHADER_STORAGE_BUFFER): skips the GL call when the
         // (id, range) already at that index matches, like the array-buffer/texture/
@@ -255,17 +416,118 @@ namespace MobileGL::MG_Backend::DirectGLES {
             Uint GetBackendVertexArrayId() const { return m_backendVAOId; }
             void Bind() const;
 
+            // Draw-path memo of SyncNeccessaryBuffers' attribute walk for this VAO: the
+            // distinct enabled-attribute buffers (deduped) and the index buffer, resolved
+            // to their backend resources once. Valid while the VAO's config version is
+            // unchanged — every attach/enable/disable/format mutation bumps it (the same
+            // invariant SyncToBackend's gate already leans on), and the VAO's attribute
+            // SharedPtrs pin each memoed frontend buffer for exactly that long, so the raw
+            // pointers cannot dangle on a hit. Per-buffer cleanliness is NOT memoed here:
+            // each hit re-checks IsBufferDrawClean (resource identity, context generation,
+            // pending ops, change serial) and falls back to EnsureBufferResource for just
+            // the dirty entries via their attribute index. The IBO entry is keyed on the
+            // slot's bound-object identity instead (its slot version is a wrapping Uint16
+            // and is not covered by the config version).
+            struct ResolvedDrawBuffers {
+                struct Entry {
+                    MG_State::GLState::BufferObject* frontend = nullptr;
+                    BufferImpl::GLESBufferResource* resource = nullptr;
+                    Uint8 attribIndex = 0;
+                };
+                Bool valid = false;
+                Uint32 configVersion = 0;
+                Uint count = 0;
+                Array<Entry, MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS> entries;
+                MG_State::GLState::BufferObject* iboFrontend = nullptr;
+                BufferImpl::GLESBufferResource* iboResource = nullptr;
+                // Buffer-mutation epoch (BufferImpl::CurrentBufferMutationEpoch) at which
+                // the LAST probe pass found every entry / the IBO clean; 0 = not stamped
+                // (epochs start at 1). While a stamp matches the pre-pass epoch read, the
+                // probes are skipped outright: any path that can dirty ANY buffer bumps
+                // the epoch (the exhaustive site list lives at the epoch declaration).
+                // The IBO stamp is only trusted together with the bound-object identity
+                // compare - the VAO's index slot can rebind with no epoch or config move.
+                Uint64 vboCleanEpoch = 0;
+                Uint64 iboCleanEpoch = 0;
+            };
+            ResolvedDrawBuffers& GetResolvedDrawBuffersMemo() { return m_resolvedDrawBuffers; }
+
+            // Memo for SyncCurrentVertexAttributeValues: which of a program's ACTIVE
+            // attribute locations lack an enabled array in this VAO (those read the
+            // context's current generic value instead of a buffer). Keyed on the VAO
+            // config version (enable/disable bumps it) and the program's active-location
+            // mask. Hosted per twin — the former function-static single entry missed on
+            // every draw once the app cycled VAOs, re-reading the cold attribute slots.
+            struct PendingAttribValueMask {
+                Bool valid = false;
+                Uint32 configVersion = 0;
+                Uint32 activeMask = 0;
+                Uint32 pendingMask = 0;
+            };
+            PendingAttribValueMask& GetPendingAttribValueMaskMemo() { return m_pendingAttribValueMask; }
+
         private:
+            ResolvedDrawBuffers m_resolvedDrawBuffers;
+            PendingAttribValueMask m_pendingAttribValueMask;
             Uint m_backendVAOId = 0;
             Array<Uint, MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS> m_clientAttributeBufferIds;
             Bool m_isInitialized = false;
             Uint16 m_syncedIndexBufferVersion = 0;
+            // Identity of the buffer the version above was stamped against. Raw and never
+            // dereferenced: the slot version is a wrapping Uint16 (see the ResolvedDrawBuffers
+            // IBO memo and the packed_pixels postmortem at BindCurrentFBO), so the version
+            // alone would read a wrapped-back count with a different buffer bound as clean.
+            const MG_State::GLState::BufferObject* m_syncedIndexBufferObject = nullptr;
+            // Aggregate gate over the per-attribute walk below: the frontend bumps its config
+            // version on every per-attribute version bump (the three Bump*Version functions are
+            // its only writers), so an unchanged config version proves every per-attribute
+            // compare in SyncToBackend would come up clean. The index-buffer slot has its own
+            // version and is NOT covered. The Bool (not a sentinel value) marks "never synced".
+            Bool m_hasSyncedConfigVersion = false;
+            Uint32 m_syncedConfigVersion = 0;
             Array<MG_State::GLState::VertexAttributeVersion, MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS>
                 m_syncedAttributeVersions;
+            // Byte shift currently baked into the instanced arrays' offsets by the baseInstance
+            // emulation (see SetPendingFetchBaseInstance). It is draw state, not VAO state, so it
+            // is deliberately NOT covered by the config version: the frontend never bumps for it.
+            // Kept here because it describes what was last EMITTED, which is what the next sync
+            // has to correct.
+            Uint32 m_syncedFetchBaseInstance = 0;
+            // BufferImpl::g_bufferBackendIdGeneration as of this twin's last emit. A
+            // mismatch means some live buffer's driver id was re-minted since; the ids
+            // baked into the driver VAO's attribute/element bindings may be dead even
+            // though every frontend version matches, so the next sync re-emits them all.
+            Uint64 m_syncedBufferIdGeneration = 0;
         };
 
         extern StateBackendObjectRegistry<MG_State::GLState::VertexArrayObject, BackendVertexArrayObject>
             g_backendVertexArrayObjects;
+
+        // Shadowed glBindVertexArray: every backend VAO bind goes through here so a
+        // draw's second bind of the same VAO (SyncToBackend, then PrepareForDraw's
+        // re-bind) reaches the driver once. Invalidate whenever the ES context is
+        // replaced - ids restart and the resting binding is 0 again.
+        void BindBackendVAOId(Uint id);
+        void InvalidateVAOBindingCache();
+        // ES resets the binding to 0 when the currently bound VAO is deleted.
+        void NoteVAOIdDeleted(Uint id);
+
+        // baseInstance emulation for drivers without GL_EXT_base_instance. GL fetches an
+        // instanced array at element "floor(instance / divisor) + baseInstance", and ES has no
+        // way to say the "+ baseInstance" part - so it is folded into the attribute's own byte
+        // offset (baseInstance * stride) for every divisor'd array, which is exactly equivalent.
+        // Must be set BEFORE PrepareForDraw so the VAO sync sees it, and cleared after the draw
+        // so the next one refetches from element 0; ScopedFetchBaseInstance does both.
+        void SetPendingFetchBaseInstance(Uint32 baseInstance);
+        Uint32 GetPendingFetchBaseInstance();
+
+        class ScopedFetchBaseInstance {
+        public:
+            explicit ScopedFetchBaseInstance(Uint32 baseInstance) { SetPendingFetchBaseInstance(baseInstance); }
+            ~ScopedFetchBaseInstance() { SetPendingFetchBaseInstance(0); }
+            ScopedFetchBaseInstance(const ScopedFetchBaseInstance&) = delete;
+            ScopedFetchBaseInstance& operator=(const ScopedFetchBaseInstance&) = delete;
+        };
     } // namespace VertexArrayImpl
 
     namespace TextureImpl {
@@ -327,6 +589,25 @@ namespace MobileGL::MG_Backend::DirectGLES {
             return target == TextureTarget::Texture3D || target == TextureTarget::TextureCubeMap;
         }
 
+        // Components per texel the frontend format's client data carries, for the three-channel
+        // formats that can be widened to a four-channel colour-renderable target; 0 for everything
+        // else. See PrepareChannelWidenedUpload.
+        Uint GetWidenableClientComponentCount(TextureInternalFormat format);
+
+        // True when a widenable format's components are integer rather than normalized, which is
+        // what decides the synthetic alpha's value: GL_RGB8I and GL_RGB8_SNORM are both uploaded
+        // as GL_BYTE, but their 1.0 is 1 and 0x7F respectively.
+        Bool IsIntegerWidenableFormat(TextureInternalFormat format);
+
+        // Repacks three-component client data as four components with an alpha of 1.0 in
+        // `uploadType`, for a format the backend widened to keep a colour attachment renderable.
+        // Returns `data` untouched when no widening applies. Pure CPU and context-free so a unit
+        // test can exercise the exact packing the driver is handed; `widenedData` is the caller's
+        // scratch buffer and has to outlive the returned pointer.
+        const void* PrepareChannelWidenedUpload(Uint componentCount, const IntVec3& texelSize, const void* data,
+                                                SizeT byteSize, GLenum uploadType, Vector<Uint8>& widenedData,
+                                                Bool integerData = false);
+
         struct StateTextureBasicInfo { // Used for tracking texture state changes
             TextureInternalFormat internalFormat = TextureInternalFormat::Unknown;
             SizeT width = 0;
@@ -364,6 +645,38 @@ namespace MobileGL::MG_Backend::DirectGLES {
             void Bind(GLenum target, Uint unit = TempTextureUnit);
             Uint GetBackendTextureId() const;
 
+            // Aggregate first-level clean gate for the per-draw trio
+            // SyncTextureParamsToBackend + SyncBuiltinSamplerToBackend +
+            // SyncMipmapsToBackend: EXACTLY the conjunction of their own early-outs
+            // (params version == synced params version; builtin-sampler version ==
+            // synced sampler version; and SyncMipmapsToBackend's cheap gate - stamped
+            // trio + content version + Mipmap storage). True means each of the three
+            // would provably return without work, so the caller may skip the calls;
+            // false only falls through to the three calls, whose own gates re-decide
+            // individually - this gate must never be MORE permissive than they are.
+            // `contextId`/`samplingGeneration` are the frontend context's current
+            // values, hoisted by the caller so a per-draw list walk reads them once
+            // instead of per texture. `t` must be the live frontend texture.
+            Bool IsDrawSyncClean(const MG_State::GLState::ITextureObject* t, Uint64 contextId,
+                                 Uint64 samplingGeneration) const {
+                if (!m_isInitialized || m_syncedShapeContextId == 0 || m_syncedShapeContextId != contextId ||
+                    m_syncedShapeGeneration != samplingGeneration) {
+                    return false;
+                }
+                const Uint16 paramsVersion = t->GetTextureParamsVersion();
+                if (m_syncedShapeParamsVersion != paramsVersion || m_syncedTextureParamsVersion != paramsVersion) {
+                    return false;
+                }
+                if (m_syncedContentVersion == 0 || m_syncedContentVersion != t->GetContentVersion()) {
+                    return false;
+                }
+                const auto& samplerObject = t->GetSamplerObject();
+                if (!samplerObject || m_syncedSamplerVersion != samplerObject->GetVersion()) {
+                    return false;
+                }
+                return t->GetStorageType() == TextureStorageType::Mipmap;
+            }
+
         private:
             void RecreateBackendTexture();
 
@@ -374,14 +687,48 @@ namespace MobileGL::MG_Backend::DirectGLES {
             Bool m_isInitialized = false;
             Bool m_imageBindableStorageRequired = false;
             Bool m_backendStorageImmutable = false;
+            // Latches the "this driver has no buffer textures" report to once per texture. The
+            // report is emitted from the respecify path, which bails before recording the state
+            // it was asked to apply - so without the latch the texture stays permanently dirty
+            // and every draw of every frame logs the same line.
+            Bool m_bufferTextureUnsupportedReported = false;
             StateTextureBasicInfo m_prevTextureInfo;
+            // Frontend content version at the last completed mipmap sync. The per-draw
+            // clean probe compares this before rebuilding shape info and scanning
+            // per-level dirty flags; 0 never matches a real version (they start at 1).
+            Uint64 m_syncedContentVersion = 0;
+            // First-level clean gate for SyncMipmapsToBackend, checked before even the
+            // IsComplete()/shape-probe walk. Valid only as a trio with the content and
+            // texture-params versions: the context's sampling-resolution generation moves on
+            // EVERY texture-shape mutation (BumpShapeVersion is the only writer of shape and
+            // unconditionally bumps it), the content version on every CPU pixel mutation, and
+            // the params version covers SetSamples/SetFixedSampleLocations, which bump neither
+            // of the other two but feed the shape probe. The context id pins the generation to
+            // the context that produced it - generations restart at 0 with a new context, and a
+            // texture is owned by exactly one context (share groups are not implemented), so a
+            // mutation can never happen under a context this key does not name. 0 = never
+            // stamped (real context ids start at 1). Backend-side invalidation rides on
+            // m_isInitialized: RequireImageBindableStorage and RecreateBackendTexture clear it.
+            Uint64 m_syncedShapeContextId = 0;
+            Uint64 m_syncedShapeGeneration = 0;
+            Uint16 m_syncedShapeParamsVersion = 0;
             SamplerParameters m_cacheSamplerParameters;
             UintVec2 m_cacheLodRange = {0, 1000};
             FloatVec4 m_cacheBorderColor = {0.0f, 0.0f, 0.0f, 0.0f};
             Vec4<TextureSwizzleParam> m_cacheSwizzleParams = {TextureSwizzleParam::Red, TextureSwizzleParam::Green,
                                                               TextureSwizzleParam::Blue, TextureSwizzleParam::Alpha};
+            // GL_DEPTH_STENCIL_TEXTURE_MODE. GL_DEPTH_COMPONENT is the GL and ES default, so a
+            // texture that never asks for the stencil aspect never emits the call. The
+            // depth/stencil readback and replicate-blit emulations also write this parameter
+            // raw, but only ever on their own scratch textures (never on an application
+            // texture), so they cannot desynchronise this cache.
+            GLenum m_cacheDepthStencilTextureMode = GL_DEPTH_COMPONENT;
             Uint16 m_syncedSamplerVersion = 0;
             Uint16 m_syncedTextureParamsVersion = 0;
+            // Set when the driver texture underneath was regenerated and has therefore lost every
+            // parameter already pushed onto it: the params-version early-out has to be overridden
+            // once, or an unchanged version would skip the re-push forever.
+            Bool m_forceTextureParamsResync = false;
         };
 
         void ActivateTextureUnit(Uint unit);
@@ -391,19 +738,28 @@ namespace MobileGL::MG_Backend::DirectGLES {
         SharedPtr<BackendTextureObject>& SyncTextureObjectToBackend(
             const SharedPtr<MG_State::GLState::ITextureObject>& textureObject,
             Bool imageBindableStorageRequired = false);
+        // Brings every texture the next draw reads - the touched units' bindings and the draw
+        // FBO's texture attachments - onto the backend, through the two borrowed-pair memos
+        // documented at their definitions. Declared here so tests can drive those memos directly.
+        void SyncNeccessaryTextures();
         extern Array<Array<BackendTextureObject*, (SizeT)TextureTarget::TextureTargetCount>,
                      MG_State::GLState::TextureState::MAX_TEXTURE_IMAGE_UNITS>
             g_boundTexturesCache;
         extern Uint g_activeTextureUnit;
-        // Bumped when the backend ES context is destroyed; texture ids stamped with
-        // an older generation belong to a dead context and must not be deleted.
-        extern Uint g_textureContextGeneration;
     } // namespace TextureImpl
 
     namespace FramebufferImpl {
         class BackendFramebufferObject {
         public:
             BackendFramebufferObject();
+            // Deletes the driver framebuffer and scrubs the binding shadow. Without it every
+            // frontend glDeleteFramebuffers leaked one ES framebuffer for the process lifetime;
+            // an app that creates a framebuffer per readback (GL CTS packed_pixels does ~3300
+            // per case) walked the driver into hundreds of megabytes of dead framebuffers and
+            // out of the resources a later attachment needs.
+            ~BackendFramebufferObject();
+            BackendFramebufferObject(const BackendFramebufferObject&) = delete;
+            BackendFramebufferObject& operator=(const BackendFramebufferObject&) = delete;
             void SyncToBackend(const SharedPtr<MG_State::GLState::FramebufferObject>& stateFBOObject,
                                FramebufferTarget asTarget);
             // Apply only this FBO's read buffer (glReadBuffer) to the backend. Split out so it can
@@ -418,6 +774,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
 
         private:
             Uint m_backendFBOId = 0;
+            Uint m_contextGeneration = 0;
 
             /* this will save buffers in its original form,
                reversion, absence or not consecutive are all allowed, as long as GL spec allows it
@@ -433,11 +790,38 @@ namespace MobileGL::MG_Backend::DirectGLES {
                this array could be provided as data directly to ES `glDrawBuffers` function
              */
             GLenum m_backendDrawBuffers[MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS] = {GL_NONE};
+
+            static constexpr Uint MAX_COLOR_ATTACHMENT_SLOTS =
+                static_cast<Uint>(FramebufferAttachmentType::Color31) -
+                static_cast<Uint>(FramebufferAttachmentType::Color0) + 1;
+            /* Where each frontend GL_COLOR_ATTACHMENTn image physically lives in the backend ES
+               framebuffer, as a GL_COLOR_ATTACHMENTm enum. ES only accepts glDrawBuffers bufs[s] ==
+               GL_COLOR_ATTACHMENTs, so a GL draw-buffer slot s naming attachment a forces a's image
+               under backend slot s. This table is the single owner of that decision and is kept a
+               PERMUTATION of the backend colour slots: every other attachment keeps its identity
+               slot when that slot survived, and is parked on the lowest free slot when it did not.
+               Deriving the point per-query from the draw-buffer array instead handed the identity
+               point to any attachment that was not a draw buffer - i.e. exactly the point a
+               relocated draw buffer had just taken over. The permutation is only true of the
+               PHYSICAL framebuffer because the attachment loop detaches a point whose frontend
+               owner is empty; do not remove that detach. */
+            GLenum m_backendColorSlots[MAX_COLOR_ATTACHMENT_SLOTS] = {GL_NONE};
+            /* Rebuild m_backendColorSlots from the frontend draw-buffer array. Returns true when any
+               attachment moved, i.e. when the physical attachments and the memoised read buffer have
+               to be re-applied. */
+            Bool RecomputeBackendColorSlots(
+                const MG_State::GLState::FramebufferObject::FramebufferAttachmentArray& stateDrawBuffers);
+
             FramebufferAttachmentType m_frontendReadBuffer = FramebufferAttachmentType::Color0;
             GLenum m_backendReadBuffer = GL_COLOR_ATTACHMENT0;
 
             using FramebufferObject = MG_State::GLState::FramebufferObject;
             FramebufferObject::FramebufferAttachmentVersionArray m_syncedFrontendAttachmentVersions = {0};
+            // g_attachmentBackendIdGeneration as of this twin's last attachment walk. A
+            // mismatch means some backend texture id was re-minted since, and any of this
+            // twin's attachment points may still hold the dead id even though the frontend
+            // attachment versions match - so the walk re-attaches everything first.
+            Uint64 m_syncedBackendIdGeneration = 0;
         };
 
         extern StateBackendObjectRegistry<MG_State::GLState::FramebufferObject, BackendFramebufferObject>
@@ -449,14 +833,97 @@ namespace MobileGL::MG_Backend::DirectGLES {
         // has to apply the clamp itself.
         Bool IsFixedPointFallbackReadAttachment();
 
-        extern Array<Uint16, SizeT(FramebufferTarget::FramebufferTargetCount)> g_fboBindVersions;
+        // True when the read buffer names a three-channel attachment the backend actually stores
+        // in a four-channel format (the colour-renderable widening). A format without alpha reads
+        // back as 1.0, so the readback path has to overwrite the alpha the draw left behind -
+        // unconditionally, since this is the format's own semantics rather than the
+        // GL_CLAMP_READ_COLOR rule the clamp above implements.
+        Bool IsAlphaWidenedFallbackReadAttachment();
+
+        // True when this attachment's storage carries an alpha channel its frontend format does
+        // not (the three-channel colour-renderable widening).
+        Bool IsAlphaWidenedColorAttachment(const MG_State::GLState::FramebufferAttachmentObject& attachmentObject);
+
+        // Bit i set = DRAW BUFFER i of `fbo` resolves to a colour attachment the backend widened
+        // from three channels to four. Indexed by draw-buffer slot, not by attachment point,
+        // because that is what glColorMaski / glClearBufferfv address.
+        Uint32 ComputeAlphaWidenedDrawBufferMask(const MG_State::GLState::FramebufferObject& fbo);
+
+        // The same mask for whatever is currently bound to GL_DRAW_FRAMEBUFFER, recomputed by
+        // SyncCurrentFBO (BackendFramebufferObject::SyncToBackend for the DRAW target, and reset
+        // to 0 on the default framebuffer). Read by the draw/clear state sync, so it is only
+        // trustworthy after SyncCurrentFBO has run in the same entry point.
+        //
+        // WHY IT EXISTS (the dst-alpha discipline). A widened attachment has a real alpha channel
+        // the application's format does not, and GL says a missing channel reads as 1.0. Readback
+        // can paper over that (ForceWideReadAlphaToOne), but GL_DST_ALPHA /
+        // GL_ONE_MINUS_DST_ALPHA blending and glBlitFramebuffer read the STORED alpha inside the
+        // driver where no interception is possible. So the stored alpha is kept at 1.0 instead:
+        // a clear touching a widened buffer writes alpha 1.0, and every draw into it has its
+        // alpha write mask forced off, so nothing can ever move it again. The application's own
+        // colour mask is untouched - glGet(GL_COLOR_WRITEMASK) still reports what it set.
+        extern Uint32 g_alphaWidenedDrawBufferMask;
+
+        // Bit i set = DRAW BUFFER i of the framebuffer bound as DRAW resolves to a colour
+        // attachment with an INTEGER format. Recomputed beside the mask above and for its sake:
+        // glClearBufferfv on an integer colour buffer is GL_INVALID_OPERATION, so the
+        // per-draw-buffer clear route the widening needs has to stand down when one is present.
+        // (glClear on an integer colour buffer is left undefined by ES in the first place, and
+        // an application that wants a defined answer has to call glClearBufferuiv/iv - which does
+        // carry the widened alpha substitution.)
+        extern Uint32 g_integerColorDrawBufferMask;
+
+        // The colour a clear has to hand the driver for one draw buffer: the application's value,
+        // except that a widened attachment's alpha is replaced by the 1.0 its three-channel
+        // format implies. `one` is 1.0 encoded in the clear call's own component type - the
+        // integer clears carry the integer 1, the float clear carries 1.0f.
+        //
+        // Returns `value` itself when nothing is substituted, so the ordinary path allocates and
+        // copies nothing; `scratch` is the caller's buffer and has to outlive the returned
+        // pointer. Free of GL state on purpose, so the substitution can be unit-tested exactly as
+        // the driver sees it.
+        template <typename T>
+        const T* SubstituteWidenedClearAlpha(const T* value, Bool widened, T one, T (&scratch)[4]) {
+            if (!widened || value == nullptr) {
+                return value;
+            }
+            scratch[0] = value[0];
+            scratch[1] = value[1];
+            scratch[2] = value[2];
+            scratch[3] = one;
+            return scratch;
+        }
+
+        // What SyncCurrentFBO last pushed for each target, as a (binding, object, revision)
+        // triple; it re-syncs unless all three still match. Stamped by SyncCurrentFBO and
+        // ForceBindCurrentFBO, cleared by InvalidateFramebufferBindingCache. The three are
+        // only meaningful together - see SyncCurrentFBO.
+        //
+        // The binding slot's own version, which changes whenever a different object is bound
+        // to this target. Distinguishes a rebind from an in-place edit, and keeps the raw
+        // pointer below from matching an address the allocator recycled for a new FBO.
+        extern Array<Uint16, SizeT(FramebufferTarget::FramebufferTargetCount)> g_fboSyncedSlotVersions;
         // Tracks the bound FBO's object version (bumped on any attachment/drawbuffer change)
         // per target: re-attaching textures or changing draw buffers on an already-bound FBO
         // must re-sync it even when the binding-slot version has not moved.
         extern Array<Uint16, SizeT(FramebufferTarget::FramebufferTargetCount)> g_fboSyncedObjectVersions;
+        // Which object was synced. Raw and never dereferenced: only compared for identity.
         extern Array<MG_State::GLState::FramebufferObject*, SizeT(FramebufferTarget::FramebufferTargetCount)>
             g_fboSyncedObjects;
 
+        // Bumped whenever a live backend texture's driver id is re-minted while its
+        // frontend texture may still be attached to application FBOs
+        // (BackendTextureObject::RecreateBackendTexture - e.g. a respecify of a texture
+        // whose backend storage went immutable). The FBO twins' attachment memos key on
+        // FRONTEND attachment versions, which a backend-side re-mint does not move, so
+        // the driver FBO would keep the deleted texture name attached forever. The
+        // SyncCurrentFBO gate compares this generation (below) to re-enter the sync,
+        // and each twin re-arms its per-attachment memo on a mismatch (SyncToBackend).
+        extern Uint64 g_attachmentBackendIdGeneration;
+        // What g_attachmentBackendIdGeneration was when SyncCurrentFBO last stamped each
+        // target; part of the synced tuple above.
+        extern Array<Uint64, SizeT(FramebufferTarget::FramebufferTargetCount)> g_fboSyncedBackendIdGenerations;
+
         // Driver-level READ/DRAW framebuffer-binding shadow. Every backend
         // glBindFramebuffer routes through BindFramebufferId so scoped helpers can
         // save/restore the current binding without a glGetIntegerv round-trip (that
@@ -467,6 +934,10 @@ namespace MobileGL::MG_Backend::DirectGLES {
         void BindFramebufferId(GLenum fbTarget, Uint id);
         Uint CurrentFramebufferBinding(FramebufferTarget target);
         void InvalidateFramebufferBindingCache();
+        // A driver framebuffer id is about to be deleted: ES reverts every target that
+        // currently binds it to 0, so the binding shadow has to follow or the next
+        // BindFramebufferId(0) would be deduped away and leave the deleted name bound.
+        void NoteFramebufferIdDeleted(Uint id);
     } // namespace FramebufferImpl
 
     // Shared scratch framebuffers for the readback/copy/blit emulation paths, with a
@@ -549,6 +1020,10 @@ namespace MobileGL::MG_Backend::DirectGLES {
         void InvalidatePackStateCache();
     } // namespace PixelStoreImpl
 
+    namespace SamplerImpl {
+        class BackendSamplerObject; // for PrgramImpl's sampler-pass memo rows below
+    }
+
     // Image uniforms take their unit from the layout(binding=N) qualifier baked into
     // the transpiled ESSL; unlike samplers they must not (and in ES cannot) be
     // assigned through glUniform1i.
@@ -595,6 +1070,43 @@ namespace MobileGL::MG_Backend::DirectGLES {
                 Float lastAssignedLodBias = 0.0f;
             };
 
+            // Memo of the whole per-draw sampler-uniform pass (glUniform1i unit
+            // assignments, lod-bias uniform, raw-depth-fetch substitution and the
+            // per-unit sampler-object binds) in BindCurrentProgramWithResources.
+            // The pass is a pure function of the keys below, and its only driver-side
+            // effect is the sampler binding of each sampled unit, so replaying it as
+            // "do nothing" additionally requires those bindings to still be on the
+            // driver - the per-entry row compare against g_boundSamplersCache (the
+            // shadow every sampler bind in this backend already routes through).
+            //
+            // Invalidation enumeration:
+            //  * sampler-uniform unit assignment (glUniform1i) and uniform-block
+            //    binding edits -> frontend backendStateVersion;
+            //  * any texture/sampler bind moving on any unit (incl. the high-water
+            //    mark moving) -> unitBindingsEpoch;
+            //  * any sampler parameter (incl. lod bias, compare mode) or texture
+            //    shape/format change -> samplingGeneration;
+            //  * another frontend context -> contextId (never-reused id);
+            //  * ES context recreation -> textureContextGeneration;
+            //  * relink / backend program rebuild -> SyncToBackend resets `valid`
+            //    (it rebuilds m_samplerUniformBindings, whose lastAssignedUnit /
+            //    lastAssignedLodBias dedup state this memo leans on);
+            //  * any other writer moving a sampled unit's sampler binding
+            //    (BindCurrentUnitSamplers on a unit-sampler change, scratch binds)
+            //    -> the row snapshot compare.
+            struct SamplerPassMemo {
+                static constexpr SizeT kMaxEntries = 16;
+                Bool valid = false;
+                Uint8 count = 0;
+                Uint64 contextId = 0;
+                Uint64 unitBindingsEpoch = 0;
+                Uint64 samplingGeneration = 0;
+                Uint32 backendStateVersion = 0;
+                Uint textureContextGeneration = 0;
+                Array<Uint8, kMaxEntries> units{};
+                Array<SamplerImpl::BackendSamplerObject*, kMaxEntries> rows{};
+            };
+
             BackendProgramObjectImpl();
             ~BackendProgramObjectImpl();
             void SyncToBackend(const SharedPtr<MG_State::GLState::ProgramObject>& stateProgramObject);
@@ -602,6 +1114,13 @@ namespace MobileGL::MG_Backend::DirectGLES {
             void SetBaseInstance(Uint32 baseInstance) const;
             void SetBaseInstanceWordIndex(Int32 wordIndex) const;
             void SetDrawID(Uint32 drawId) const;
+            void SetBaseVertex(Int32 baseVertex) const;
+            // True when the transpiled program kept a gl_DrawID uniform, i.e. SetDrawID
+            // actually reaches a shader read rather than being discarded.
+            Bool ReadsDrawID() const { return m_drawIdUniformLocation >= 0; }
+            // Same for gl_BaseVertex: only a program that reads it pays for the per-draw
+            // uniform write, and only such a program needs the reset after one.
+            Bool ReadsBaseVertex() const { return m_baseVertexUniformLocation >= 0; }
             Int GetIndirectParamsBinding() const { return m_indirectParamsBinding; }
             Uint GetBackendProgramId() const { return m_backendProgramId; }
             // False when the last SyncToBackend could not produce a usable program (a
@@ -612,6 +1131,11 @@ namespace MobileGL::MG_Backend::DirectGLES {
             Uint32 GetSnormFallbackClampOutputMask() const { return m_snormFallbackClampOutputMask; }
             Uint32 GetUnormFallbackClampOutputMask() const { return m_unormFallbackClampOutputMask; }
             Uint GetFragColorBroadcastCount() const { return m_fragColorBroadcastCount; }
+            // Signature of the glShaderStorageBlockBinding override set the generated ESSL was
+            // transpiled against (ES can only express a storage-block binding as the declared
+            // qualifier, so the overrides are baked into the source). A mismatch means the
+            // program is stale exactly like the clamp masks above.
+            Uint64 GetShaderStorageBlockBindingSignature() const { return m_shaderStorageBlockBindingSignature; }
 
             Bool HasGlobalUboBlock() const { return m_globalUboBackendBlockIndex >= 0; }
             const Vector<Int>& GetUniformBlockBackendIndices() const { return m_uniformBlockBackendIndices; }
@@ -623,17 +1147,52 @@ namespace MobileGL::MG_Backend::DirectGLES {
             // reflected size when the transpiled block pads differently).
             Int GetGlobalUboBackendBlockSize() const { return m_globalUboBackendBlockSize; }
             BufferImpl::UboRingAllocation& GetGlobalUboRingAllocation() { return m_globalUboRingAllocation; }
+            SamplerPassMemo& GetSamplerPassMemo() { return m_samplerPassMemo; }
             // Frontend link version this backend program (and its resource caches) was
             // built from; a mismatch means every link-derived cache here is stale.
             Uint32 GetSyncedLinkVersion() const { return m_syncedLinkVersion; }
+            // Image-uniform unit generation this backend program was GENERATED against.
+            // Separate from the link version because it is not link state: ES forbids
+            // glUniform1i on an image uniform, so RebindImageUniformsToFrontendUnits bakes the
+            // unit into the ESSL, and a program built before glUniform1i moved that unit is as
+            // stale as one built before a relink - while the sampler half, which really is
+            // re-issued per draw, needs nothing of the sort.
+            Uint32 GetSyncedImageUnitVersion() const { return m_syncedImageUnitVersion; }
+            // Whether the (unit, bound format) pairs this program's FORMAT-LESS image uniforms
+            // resolve to are still the ones its ESSL was generated against.
+            //
+            // A fourth condition of the same family as the three above, and the only one that
+            // reads live state rather than a program-side counter, because that is where the
+            // dependency actually is. GLSL ES requires a format layout qualifier on every image
+            // where desktop GLSL lets a writeonly declaration omit one, and the only correct
+            // qualifier is whatever glBindImageTexture named - so a declaration with no format
+            // is compiled against the BINDING, and a rebind to a different format makes the
+            // built program wrong. Keyed on the units the program's own images address (cached
+            // at sync, since a unit can only move by glUniform1i, which bumps the image-unit
+            // version above and forces a re-sync anyway), so the cost on a program with no
+            // format-less image - which is all but a handful - is one empty-vector test.
+            //
+            // Deliberately NOT reached from glBindImageTexture: that entry point must never
+            // trigger a build (same constraint as glShaderStorageBlockBinding). It moves the
+            // state and this comparison notices at the next Prepare, which is also what makes
+            // an image first bound AFTER link work.
+            Bool ImageUnitFormatsStillMatch() const;
+            // The value ImageUnitFormatsStillMatch() compares against, recomputed from live
+            // image-unit state. 0 when the program has no format-less image uniform.
+            Uint64 ComputeImageUnitFormatSignature() const;
 
         private:
             void CacheResourceLocations(const SharedPtr<MG_State::GLState::ProgramObject>& stateProgramObject);
 
             Uint m_backendProgramId = 0;
+            // GL name of the frontend program this was last synced from; diagnostics only, so
+            // an unusable backend program can be traced back to the glCreateProgram id the app
+            // knows it by.
+            Uint m_frontendProgramId = 0;
             Uint m_backendGlobalUBOId = 0;
             Int m_baseInstanceUniformLocation = -1;
             Int m_drawIdUniformLocation = -1;
+            Int m_baseVertexUniformLocation = -1;
             Int m_baseInstanceWordIndexUniformLocation = -1;
             Int m_indirectParamsBinding = -1;
             Uint32 m_snormFallbackClampOutputMask = 0;
@@ -641,6 +1200,8 @@ namespace MobileGL::MG_Backend::DirectGLES {
             // Draw buffers a legacy gl_FragColor write has to reach (see
             // PrgramImpl::BroadcastLegacyFragColor); 1 keeps the plain single-output shader.
             Uint m_fragColorBroadcastCount = 1;
+            // 0 is the signature of an empty override set, i.e. what almost every program has.
+            Uint64 m_shaderStorageBlockBindingSignature = 0;
             Bool m_isInitialized = false;
             Bool m_backendProgramUsable = false;
 
@@ -651,6 +1212,14 @@ namespace MobileGL::MG_Backend::DirectGLES {
             Uint32 m_lastUploadedGlobalUboVersion = ~0u;
             BufferImpl::UboRingAllocation m_globalUboRingAllocation;
             Uint32 m_syncedLinkVersion = ~0u;
+            Uint32 m_syncedImageUnitVersion = ~0u;
+            // Image units addressed by the program's FORMAT-LESS image uniforms, and the digest
+            // of the (unit, format) pairs the generated ESSL baked. Empty/0 for every program
+            // that declares a format on all of its images, which is the overwhelming majority -
+            // and what keeps the per-draw comparison free for them.
+            Vector<Int> m_formatlessImageUnits;
+            Uint64 m_imageUnitFormatSignature = 0;
+            SamplerPassMemo m_samplerPassMemo;
         };
 
         extern Uint32 g_snormFallbackClampOutputMask;
@@ -665,18 +1234,87 @@ namespace MobileGL::MG_Backend::DirectGLES {
         extern Uint g_lastUsedBackendProgramId;
         extern StateBackendObjectRegistry<MG_State::GLState::ProgramObject, BackendProgramObjectImpl>
             g_backendProgramObjects;
+
+        // Points one shader storage block of an ALREADY-LINKED backend program at
+        // `binding`. `blockName` is the frontend interface-query spelling; the real
+        // driver's own index for it is looked up here, because the transpiled ESSL's
+        // block order is not the frontend's. Returns false when the block does not exist
+        // on the backend program (eliminated as unused, or the driver lacks the entry
+        // points), which is not an error - GL_BUFFER_BINDING is served from the frontend
+        // record either way.
+        //
+        // NOT how a rebinding reaches the shader. glShaderStorageBlockBinding has no ES
+        // equivalent and is absent from every real ES driver, so this is a no-op there;
+        // SyncToBackend bakes the effective binding into the ESSL it generates instead
+        // (SpvcSession::SetShaderStorageBlockBinding). This is kept as the cheaper path on
+        // a driver that does happen to expose the entry point.
+        Bool ApplyShaderStorageBlockBinding(Uint backendProgramId, const String& blockName, Uint binding);
+        // Replays every glShaderStorageBlockBinding recorded on the program onto a backend
+        // program that was just built - best effort, on the same "only where the driver has
+        // the entry point" terms as ApplyShaderStorageBlockBinding above. Mirrors
+        // DirectVulkan's reseed-on-rebuild in BuildProgramResourceCache.
+        void ReseedShaderStorageBlockBindings(Uint backendProgramId,
+                                              const MG_State::GLState::ProgramObject& stateProgramObject);
+        // Order-independent digest of the program's glShaderStorageBlockBinding overrides.
+        // The generated ESSL carries them (ES has no way to move a storage block's binding
+        // after link), so a program built against a different set is stale and the draw path
+        // has to rebuild it. Computed from the values, so re-setting a block to the binding it
+        // already has costs nothing. 0 when nothing was ever rebound.
+        Uint64 ComputeShaderStorageBlockBindingSignature(
+            const MG_State::GLState::ProgramObject& stateProgramObject);
+
+        // Everything the image-format bake needs from one walk of a program's uniform
+        // reflection. GLSL ES requires a format layout qualifier on every image uniform;
+        // desktop GLSL lets a writeonly (or readonly) declaration omit one, and the only
+        // format that is CORRECT to substitute is whatever glBindImageTexture named for the
+        // unit that uniform addresses - so the transpile bakes it in and the build is keyed
+        // on it.
+        struct ImageFormatBakeInputs {
+            // Uniform name (SPIR-V spelling, i.e. an array named once, unsubscripted) to the GL
+            // internal format to bake. Holds only uniforms that DECLARED no format; a declared
+            // one is authoritative and is never overridden.
+            UnorderedMap<String, Uint> glFormatByUniformName;
+            // The same uniforms whose format SPIRV-Cross REFUSES to print for ESSL (it throws on
+            // its desktop-only set, which loses the stage), paired with the ESSL spelling to
+            // write into the emitted declaration instead. Disjoint from the map above by
+            // construction: a format is baked into the module or completed in the text, never
+            // both. r8ui - the stencil half of the packed_depth_stencil case - lands here.
+            UnorderedMap<String, String> esslFormatQualifierByUniformName;
+            // Units those uniforms address, kept so the draw path can re-read their formats
+            // without walking the reflection again.
+            Vector<Int> units;
+            // Digest of the (unit, format) pairs above. 0 when the program has no format-less
+            // image uniform, which is all but a handful.
+            Uint64 signature = 0;
+            // Array uniforms whose elements resolved to units holding DIFFERENT formats: one
+            // declaration carries one qualifier, so there is nothing correct to bake and they
+            // are dropped from the map above. Kept for diagnostics.
+            Vector<String> conflictedNames;
+            // Some format in play - declared or baked - is outside the GLSL ES core image
+            // format set, so the emitted ESSL needs the GL_NV_image_formats directive.
+            Bool needsExtendedImageFormats = false;
+        };
+        ImageFormatBakeInputs CollectImageFormatBakeInputs(
+            const MG_State::GLState::ProgramObject& stateProgramObject);
     } // namespace PrgramImpl
 
     namespace SamplerImpl {
         class BackendSamplerObject {
         public:
             BackendSamplerObject();
+            // Deletes the driver sampler and clears the units whose binding shadow still names
+            // this twin (a recycled heap address would otherwise false-skip a later Bind).
+            // Frontend glDeleteSamplers used to leak the backend id for the process lifetime.
+            ~BackendSamplerObject();
+            BackendSamplerObject(const BackendSamplerObject&) = delete;
+            BackendSamplerObject& operator=(const BackendSamplerObject&) = delete;
             void SyncToBackend(const SharedPtr<MG_State::GLState::SamplerObject>& stateSamplerObject);
             void Bind(Uint unit);
             Uint GetBackendSamplerId() const;
 
         private:
             Uint m_backendSamplerId = 0;
+            Uint m_contextGeneration = 0;
             Bool m_isInitialized = false;
             SamplerParameters m_cacheSamplerParameters;
             Uint16 m_syncedSamplerVersion = 0;
@@ -694,12 +1332,18 @@ namespace MobileGL::MG_Backend::DirectGLES {
         class BackendRenderbufferObject {
         public:
             BackendRenderbufferObject();
+            // Deletes the driver renderbuffer; frontend glDeleteRenderbuffers used to leak it
+            // (with its whole image allocation) for the process lifetime.
+            ~BackendRenderbufferObject();
+            BackendRenderbufferObject(const BackendRenderbufferObject&) = delete;
+            BackendRenderbufferObject& operator=(const BackendRenderbufferObject&) = delete;
             void SyncToBackend(const SharedPtr<MG_State::GLState::RenderbufferObject>& stateRBOObject);
             Uint GetBackendRenderbufferId() const { return m_backendRBOId; }
             void Bind() const;
 
         private:
             Uint m_backendRBOId = 0;
+            Uint m_contextGeneration = 0;
             Bool m_isInitialized = false;
             TextureInternalFormat m_cacheInternalFormat = TextureInternalFormat::Unknown;
             Int m_cacheWidth = 0;

MobileGL/MG_Backend/DirectGLES/MultiDraw.cpp

@@ -0,0 +1,928 @@
+// MobileGL - MobileGL/MG_Backend/DirectGLES/MultiDraw.cpp
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#include "MultiDraw.h"
+#include "Managers.h"
+#include <MG_State/GLState/Core.h>
+#include <cstring>
+#include <limits>
+
+namespace MobileGL::MG_Backend::DirectGLES::MultiDrawImpl {
+    using MG_Config::GLESMultiDrawMode;
+
+    namespace {
+        // ---------------------------------------------------------------------------
+        // Batch shape
+        // ---------------------------------------------------------------------------
+
+        SizeT IndexTypeSize(GLenum type) {
+            switch (type) {
+            case GL_UNSIGNED_BYTE: return 1;
+            case GL_UNSIGNED_SHORT: return 2;
+            case GL_UNSIGNED_INT: return 4;
+            default: return 0;
+            }
+        }
+
+        // The all-ones value of an index type, which is what GL restarts on once
+        // primitive restart is in play. CheckPrimitiveRestartSupported has already
+        // rejected the arbitrary-index form of GL_PRIMITIVE_RESTART, so an enabled
+        // restart always restarts here and nowhere else.
+        Uint32 RestartSentinelFor(GLenum type) {
+            switch (type) {
+            case GL_UNSIGNED_BYTE: return 0xFFu;
+            case GL_UNSIGNED_SHORT: return 0xFFFFu;
+            default: return 0xFFFFFFFFu;
+            }
+        }
+
+        Bool RestartActive() {
+            return MG_State::pGLContext->IsCapabilityEnabled(CapabilityInput::PrimitiveRestart) ||
+                   MG_State::pGLContext->IsCapabilityEnabled(CapabilityInput::PrimitiveRestartFixedIndex);
+        }
+
+        // Vertices per primitive for the modes whose sub-draws may be concatenated into a
+        // single draw without changing the primitive stream. Zero for strip/loop/fan modes
+        // (concatenation would weld one sub-draw's last primitive to the next sub-draw's
+        // first) and for GL_PATCHES, whose primitive size is dynamic tessellation state.
+        Uint32 ConcatenablePrimitiveSize(GLenum mode) {
+            switch (mode) {
+            case GL_POINTS: return 1;
+            case GL_LINES: return 2;
+            case GL_TRIANGLES: return 3;
+            case GL_LINES_ADJACENCY: return 4;
+            case GL_TRIANGLES_ADJACENCY: return 6;
+            default: return 0;
+            }
+        }
+
+        // Beyond this an emulated batch would ask for a scratch allocation measured in
+        // hundreds of megabytes (and the scratch ring never shrinks again); decline and let
+        // a per-sub-draw tier handle it instead of trying and failing inside the driver.
+        constexpr SizeT kMaxFlattenedIndices = SizeT{1} << 24;
+
+        // The flattening dispatch is one invocation per output index. ES 3.1 only
+        // guarantees 65535 work groups per dimension, and exceeding it makes
+        // glDispatchCompute an INVALID_VALUE no-op - which would leave the draw reading an
+        // uninitialised index buffer rather than failing visibly. Cap the tier there
+        // instead of querying: 4.19M indices is far past any real multi-draw batch, and
+        // beyond it the per-sub-draw tiers are the better answer anyway.
+        constexpr SizeT kComputeWorkGroupSize = 64;
+        constexpr SizeT kMaxComputeWorkGroups = 65535;
+        constexpr SizeT kMaxComputeFlattenedIndices = kMaxComputeWorkGroups * kComputeWorkGroupSize;
+
+        Uint BoundDrawIndirectBufferId() {
+            const auto& indirect =
+                MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DrawIndirect).GetBoundObject();
+            if (!indirect) return 0;
+            const auto* resource = BufferImpl::EnsureBufferResource(indirect);
+            return resource ? resource->id : 0;
+        }
+
+        const SharedPtr<MG_State::GLState::BufferObject>& BoundIndexBuffer() {
+            static const SharedPtr<MG_State::GLState::BufferObject> none;
+            const auto& vao = MG_State::pGLContext->GetBoundVertexArray();
+            if (!vao) return none;
+            return vao->GetIndexBufferBindingSlot().GetBoundObject();
+        }
+
+        // The GL name PrepareForDraw left on GL_ELEMENT_ARRAY_BUFFER, i.e. what a tier
+        // that swaps in a scratch index buffer has to put back. Restoring the exact name
+        // matters beyond tidiness: the VAO twin memoises that it already synced this
+        // index binding and will not re-issue it on the next draw.
+        Uint BoundIndexBufferId() {
+            const auto& ibo = BoundIndexBuffer();
+            if (!ibo) return 0;
+            const auto* resource = BufferImpl::EnsureBufferResource(ibo);
+            return resource ? resource->id : 0;
+        }
+
+        // ---------------------------------------------------------------------------
+        // Scratch GL objects
+        //
+        // All of them belong to the ES context and are abandoned (not deleted) when it
+        // dies, exactly like XfbImpl's scatter buffer: the names are the dead context's
+        // to reclaim, and deleting them would target whatever the successor context
+        // handed out for the same name.
+        // ---------------------------------------------------------------------------
+
+        struct ScratchBuffer {
+            Uint id = 0;
+            SizeT capacity = 0;
+            SizeT cursor = 0; // ring buffers only: next free byte
+        };
+
+        ScratchBuffer g_indirectCommands; // synthesized DrawElementsIndirectCommand array
+        ScratchBuffer g_rebasedIndices;   // CPU-rebased index stream
+        ScratchBuffer g_drawInfo;         // compute tier: per-sub-draw descriptors
+        ScratchBuffer g_flattenedIndices; // compute tier: flattened index stream
+
+        Uint g_computeProgram = 0;
+        Bool g_computeProgramFailed = false;
+        GLint g_uElementSize = -1;
+        GLint g_uDrawCount = -1;
+        GLint g_uTotalIndices = -1;
+
+        // Reused staging, so a steady stream of batches allocates nothing.
+        Vector<DrawElementsIndirectCommand> g_commandStaging;
+        Vector<Uint32> g_indexStaging;
+        Vector<Uint32> g_drawInfoStaging;
+        Vector<GLint> g_zeroBaseVertices;
+
+        // Everything below stages through GL_ARRAY_BUFFER, the manager-wide staging target
+        // (BufferImpl::TempBufferTarget); binding it disturbs no VAO state.
+        Bool EnsureScratchName(ScratchBuffer& buffer) {
+            if (buffer.id != 0) return true;
+            GLuint id = 0;
+            g_GLESFuncs.glGenBuffers(1, &id);
+            if (id == 0) return false;
+            buffer.id = id;
+            buffer.capacity = 0;
+            buffer.cursor = 0;
+            return true;
+        }
+
+        // Whole-buffer upload, for the two buffers that are read from offset 0 because they
+        // are bound as storage blocks. Respecifies rather than sub-updates: glBufferData
+        // orphans the previous store, so the upload never waits on a dispatch still reading
+        // the old contents out of the same name.
+        Bool UploadScratch(ScratchBuffer& buffer, SizeT bytes, const void* data) {
+            if (bytes == 0) return true;
+            if (!EnsureScratchName(buffer)) return false;
+            BufferImpl::BindBufferId(BufferImpl::TempBufferTarget, buffer.id);
+            // Grow in powers of two so a batch that creeps up in size stops respecifying.
+            SizeT capacity = buffer.capacity == 0 ? bytes : buffer.capacity;
+            while (capacity < bytes) capacity *= 2;
+            g_GLESFuncs.glBufferData(BufferImpl::TempBufferTarget, static_cast<GLsizeiptr>(capacity), nullptr,
+                                     GL_STREAM_DRAW);
+            buffer.capacity = capacity;
+            buffer.cursor = 0;
+            if (data) {
+                g_GLESFuncs.glBufferSubData(BufferImpl::TempBufferTarget, 0, static_cast<GLsizeiptr>(bytes), data);
+            }
+            return true;
+        }
+
+        // Ring upload, for the buffers whose consumers can address a byte offset (indirect
+        // commands and rewritten index streams). Respecifying per batch is what an
+        // orphan-every-time scheme costs, and on a desktop-class driver that allocation
+        // dominated the tiers that use these buffers - a multi-draw of 32 sub-draws stages
+        // 640 bytes and paid for a fresh store to hold them. Bump-allocating instead means
+        // one respecify per wrap; every byte between two wraps is written exactly once, so
+        // nothing in flight is overwritten, and the wrap itself orphans.
+        constexpr SizeT kRingAlignment = 16; // >= 4, so both command and uint32-index offsets stay legal
+        constexpr SizeT kMinRingBytes = 1u << 16;
+
+        Bool UploadScratchRing(ScratchBuffer& buffer, SizeT bytes, const void* data, SizeT& outOffset) {
+            outOffset = 0;
+            if (bytes == 0) return true;
+            if (!EnsureScratchName(buffer)) return false;
+            BufferImpl::BindBufferId(BufferImpl::TempBufferTarget, buffer.id);
+
+            const SizeT aligned = (bytes + kRingAlignment - 1) & ~(kRingAlignment - 1);
+            if (buffer.capacity < aligned) {
+                SizeT capacity = buffer.capacity == 0 ? kMinRingBytes : buffer.capacity;
+                while (capacity < aligned) capacity *= 2;
+                g_GLESFuncs.glBufferData(BufferImpl::TempBufferTarget, static_cast<GLsizeiptr>(capacity), nullptr,
+                                         GL_STREAM_DRAW);
+                buffer.capacity = capacity;
+                buffer.cursor = 0;
+            } else if (buffer.cursor + aligned > buffer.capacity) {
+                g_GLESFuncs.glBufferData(BufferImpl::TempBufferTarget, static_cast<GLsizeiptr>(buffer.capacity),
+                                         nullptr, GL_STREAM_DRAW);
+                buffer.cursor = 0;
+            }
+
+            outOffset = buffer.cursor;
+            if (data) {
+                g_GLESFuncs.glBufferSubData(BufferImpl::TempBufferTarget, static_cast<GLintptr>(outOffset),
+                                            static_cast<GLsizeiptr>(bytes), data);
+            }
+            buffer.cursor += aligned;
+            return true;
+        }
+
+        // ---------------------------------------------------------------------------
+        // Tier resolution
+        // ---------------------------------------------------------------------------
+
+        // Best-first, and measured rather than assumed. MobileGlues orders its own Auto
+        // multiindirect -> indirect -> basevertex; on both ES drivers available here that
+        // is backwards, because staging a command buffer per batch costs more than the
+        // driver entries it saves. mc_sodium_multidraw (132 batches x 32 sub-draws),
+        // ns/op, median of three:
+        //
+        //                    NVIDIA ES 3.2   Mesa llvmpipe ES 3.2
+        //   ext                  n/a               19300
+        //   basevertex           2500              25200
+        //   multiindirect        5700              27600
+        //   drawelements         5600              28700
+        //   indirect             5800              31000
+        //
+        // Ring-allocating the command staging (instead of respecifying per batch) was
+        // tried first and moved the indirect tiers by less than noise, so the cost is the
+        // indirect draw path itself, not the upload. Only "ext" - a real multi-draw entry
+        // point rather than an indirect one - actually beats replaying the sub-draws.
+        //
+        // The compute tier is deliberately absent from the ladder: it rewrites the
+        // primitive stream rather than replaying it, and it measured slowest of all here,
+        // so it stays opt-in behind the env knob (the same call MobileGlues makes - its
+        // Auto never selects Compute either).
+        constexpr GLESMultiDrawMode kAutoLadder[] = {
+            GLESMultiDrawMode::Ext,      GLESMultiDrawMode::BaseVertex,   GLESMultiDrawMode::MultiIndirect,
+            GLESMultiDrawMode::Indirect, GLESMultiDrawMode::DrawElements,
+        };
+
+        Bool SupportsTier(GLESMultiDrawMode tier) {
+            return IsTierSupported(g_GLESCapabilities, g_GLESFuncs, tier);
+        }
+
+        GLESMultiDrawMode g_resolvedTier = GLESMultiDrawMode::Auto;
+        Bool g_tierResolved = false;
+        String g_tierResolution;
+
+        void ResolveTierOnce() {
+            if (g_tierResolved) return;
+            g_tierResolved = true;
+            g_resolvedTier =
+                ResolveTier(g_GLESCapabilities, g_GLESFuncs, MG_Config::Features.EsprytMultiDrawMode,
+                            &g_tierResolution);
+            MGLOG_D("DirectGLES multi-draw: %s", g_tierResolution.c_str());
+        }
+
+        // Which tiers have already announced themselves, one bit per GLESMultiDrawMode.
+        // The resolution line above says which tier was CHOSEN; this says which one a
+        // batch actually went through, and the two differ whenever a batch's shape
+        // demotes it. Worth a line each: a multi-draw path that resolves to a tier and
+        // then quietly runs a different one is exactly how "the batch drew nothing"
+        // hides.
+        Uint32 g_announcedTiers = 0;
+
+        void NoteTierExecuted(GLESMultiDrawMode tier) {
+            const Uint32 bit = 1u << static_cast<Uint32>(tier);
+            if (g_announcedTiers & bit) return;
+            g_announcedTiers |= bit;
+            MGLOG_D("DirectGLES multi-draw: first batch executed via tier \"%s\"", TierName(tier));
+        }
+
+        // The tier this particular batch can actually take. A tier is demoted here when
+        // the batch's own shape - not the driver - rules it out; the compute tier keeps
+        // its remaining feasibility checks inside its implementation, where the data it
+        // has to walk is already in hand.
+        GLESMultiDrawMode ResolveTierForBatch(Bool programReadsDrawID, Bool perSubDrawBaseVertex,
+                                              Bool hasIndexBuffer) {
+            ResolveTierOnce();
+            GLESMultiDrawMode tier = g_resolvedTier;
+
+            // Batched tiers issue one driver entry for the whole batch, so the emulated
+            // gl_DrawID uniform can only hold one value across every sub-draw. A program
+            // that reads gl_DrawID gets an unrolled tier, which feeds each sub-draw its
+            // own index (the spec's value); nothing else observes the difference. The
+            // emulated gl_BaseVertex is one uniform for the same reason, so a batch whose
+            // sub-draws carry their own base vertices unrolls too - even the Ext tier,
+            // which hands the driver the whole basevertex array, can only leave ONE value
+            // in the uniform the shader reads.
+            const Bool batched = tier == GLESMultiDrawMode::Ext || tier == GLESMultiDrawMode::MultiIndirect ||
+                                 tier == GLESMultiDrawMode::Compute;
+            if (batched && (programReadsDrawID || perSubDrawBaseVertex)) {
+                tier = SupportsTier(GLESMultiDrawMode::BaseVertex) ? GLESMultiDrawMode::BaseVertex
+                                                                   : GLESMultiDrawMode::DrawElements;
+            }
+
+            // The indirect tiers describe each sub-draw as an element offset into the
+            // bound element array buffer. A client-memory index array has no such buffer,
+            // and indirect draws are not defined without one.
+            if (!hasIndexBuffer &&
+                (tier == GLESMultiDrawMode::MultiIndirect || tier == GLESMultiDrawMode::Indirect)) {
+                tier = SupportsTier(GLESMultiDrawMode::BaseVertex) ? GLESMultiDrawMode::BaseVertex
+                                                                   : GLESMultiDrawMode::DrawElements;
+            }
+            return tier;
+        }
+
+        // ---------------------------------------------------------------------------
+        // Index rewriting, shared by the two tiers that fold base vertices into indices
+        // ---------------------------------------------------------------------------
+
+        // Both of those tiers emit GL_UNSIGNED_INT regardless of the source type. Keeping
+        // the source width would be wrong, not merely tight: GL adds baseVertex to the
+        // index at full precision, so a GL_UNSIGNED_SHORT index plus a base vertex past
+        // 65535 addresses a vertex the source type cannot spell. Widening also gives the
+        // rewritten stream a restart sentinel (0xFFFFFFFF) that survives the rebase.
+        void RebaseIndices(const Uint8* source, SizeT sourceIndexCount, SizeT indexSize, Int32 baseVertex,
+                           Bool restartActive, Uint32 restartSentinel, Uint32* out) {
+            const Uint32 baseVertexBits = static_cast<Uint32>(baseVertex);
+            for (SizeT i = 0; i < sourceIndexCount; ++i) {
+                Uint32 value = 0;
+                switch (indexSize) {
+                case 1: value = source[i]; break;
+                case 2: {
+                    Uint16 narrow = 0;
+                    std::memcpy(&narrow, source + i * 2, sizeof(narrow));
+                    value = narrow;
+                    break;
+                }
+                default: std::memcpy(&value, source + i * 4, sizeof(value)); break;
+                }
+                // Unsigned wraparound is the defined behaviour for a negative base vertex.
+                out[i] = (restartActive && value == restartSentinel) ? 0xFFFFFFFFu : value + baseVertexBits;
+            }
+        }
+
+        // CPU-readable bytes of one sub-draw's indices, from the frontend shadow of the
+        // bound index buffer or straight from the client array. Null when the sub-draw
+        // would read outside the buffer.
+        const Uint8* ResolveSubDrawIndices(const SharedPtr<MG_State::GLState::BufferObject>& indexBuffer,
+                                           const Uint8* indexBufferBytes, SizeT indexBufferSize, const void* indices,
+                                           SizeT indexCount, SizeT indexSize) {
+            if (!indexBuffer) {
+                return static_cast<const Uint8*>(indices);
+            }
+            if (!indexBufferBytes) return nullptr;
+            const SizeT byteOffset = reinterpret_cast<SizeT>(indices);
+            const SizeT byteEnd = byteOffset + indexCount * indexSize;
+            if (byteEnd > indexBufferSize || byteEnd < byteOffset) return nullptr;
+            return indexBufferBytes + byteOffset;
+        }
+
+        // ---------------------------------------------------------------------------
+        // Tier: Ext - one glMultiDrawElementsBaseVertexEXT
+        // ---------------------------------------------------------------------------
+
+        Bool RunExt(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices, GLsizei drawcount,
+                    const GLint* basevertex) {
+            if (!SupportsTier(GLESMultiDrawMode::Ext)) return false;
+            const GLint* baseVertices = basevertex;
+            if (!baseVertices) {
+                // glMultiDrawElements: every base vertex is 0, but the entry point still
+                // wants an array. One permanently-zero vector serves every such batch.
+                if (g_zeroBaseVertices.size() < static_cast<SizeT>(drawcount)) {
+                    g_zeroBaseVertices.resize(static_cast<SizeT>(drawcount), 0);
+                }
+                baseVertices = g_zeroBaseVertices.data();
+            }
+            g_GLESFuncs.glMultiDrawElementsBaseVertexEXT(mode, count, type, indices, drawcount, baseVertices);
+            NoteTierExecuted(GLESMultiDrawMode::Ext);
+            return true;
+        }
+
+        // ---------------------------------------------------------------------------
+        // Tiers: MultiIndirect / Indirect - synthesized indirect commands
+        // ---------------------------------------------------------------------------
+
+        Bool RunIndirect(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
+                         GLsizei drawcount, const GLint* basevertex, Bool batched, Bool feedDrawID,
+                         Bool feedBaseVertex) {
+            if (!SupportsTier(batched ? GLESMultiDrawMode::MultiIndirect : GLESMultiDrawMode::Indirect)) return false;
+            const SizeT indexSize = IndexTypeSize(type);
+            if (indexSize == 0) return false;
+            // Indirect commands address indices as an element offset into the bound element
+            // array buffer, and an indirect draw is not defined without one.
+            const auto& indexBuffer = BoundIndexBuffer();
+            if (!indexBuffer) return false;
+
+            g_commandStaging.resize(static_cast<SizeT>(drawcount));
+            for (GLsizei i = 0; i < drawcount; ++i) {
+                const SizeT byteOffset = reinterpret_cast<SizeT>(indices[i]);
+                // firstIndex counts elements, so an offset that is not a whole number of
+                // them cannot be expressed as a command at all.
+                if (byteOffset % indexSize != 0) return false;
+                auto& command = g_commandStaging[static_cast<SizeT>(i)];
+                command.count = count[i] > 0 ? static_cast<Uint32>(count[i]) : 0u;
+                command.instanceCount = 1;
+                command.firstIndex = static_cast<Uint32>(byteOffset / indexSize);
+                command.baseVertex = basevertex ? basevertex[i] : 0;
+                command.baseInstance = 0;
+            }
+
+            const SizeT commandBytes = g_commandStaging.size() * sizeof(DrawElementsIndirectCommand);
+            SizeT commandBase = 0;
+            if (!UploadScratchRing(g_indirectCommands, commandBytes, g_commandStaging.data(), commandBase)) {
+                return false;
+            }
+
+            // Every synthesized command carries baseInstance 0. Say so through the direct
+            // path, which also clears the indirect-params word index a preceding real
+            // indirect draw may have left pointing into its own command buffer.
+            SetCurrentBaseInstance(0);
+
+            const Uint previousIndirectBinding = BoundDrawIndirectBufferId();
+            BufferImpl::BindBufferId(GL_DRAW_INDIRECT_BUFFER, g_indirectCommands.id);
+            if (batched) {
+                g_GLESFuncs.glMultiDrawElementsIndirectEXT(mode, type, reinterpret_cast<const void*>(commandBase),
+                                                           drawcount, 0);
+            } else {
+                for (GLsizei i = 0; i < drawcount; ++i) {
+                    if (feedDrawID) SetCurrentDrawID(static_cast<Uint32>(i));
+                    if (feedBaseVertex) SetCurrentBaseVertex(basevertex ? basevertex[i] : 0);
+                    const SizeT commandOffset = commandBase + static_cast<SizeT>(i) * sizeof(DrawElementsIndirectCommand);
+                    g_GLESFuncs.glDrawElementsIndirect(mode, type, reinterpret_cast<const void*>(commandOffset));
+                }
+                if (feedDrawID) SetCurrentDrawID(0);
+                if (feedBaseVertex) SetCurrentBaseVertex(0);
+            }
+            BufferImpl::BindBufferId(GL_DRAW_INDIRECT_BUFFER, previousIndirectBinding);
+            NoteTierExecuted(batched ? GLESMultiDrawMode::MultiIndirect : GLESMultiDrawMode::Indirect);
+            return true;
+        }
+
+        // ---------------------------------------------------------------------------
+        // Tier: BaseVertex - the per-sub-draw replay
+        // ---------------------------------------------------------------------------
+
+        Bool RunBaseVertexLoop(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
+                               GLsizei drawcount, const GLint* basevertex, Bool feedDrawID, Bool feedBaseVertex) {
+            if (!SupportsTier(GLESMultiDrawMode::BaseVertex)) return false;
+            for (GLsizei i = 0; i < drawcount; ++i) {
+                if (count[i] <= 0) continue;
+                if (feedDrawID) SetCurrentDrawID(static_cast<Uint32>(i));
+                if (feedBaseVertex) SetCurrentBaseVertex(basevertex ? basevertex[i] : 0);
+                g_GLESFuncs.glDrawElementsBaseVertex(mode, count[i], type, indices[i],
+                                                     basevertex ? basevertex[i] : 0);
+            }
+            if (feedDrawID) SetCurrentDrawID(0);
+            if (feedBaseVertex) SetCurrentBaseVertex(0);
+            NoteTierExecuted(GLESMultiDrawMode::BaseVertex);
+            return true;
+        }
+
+        // ---------------------------------------------------------------------------
+        // Tier: DrawElements - base vertices folded into a scratch index stream
+        // ---------------------------------------------------------------------------
+
+        Bool RunRebasedDrawElements(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
+                                    GLsizei drawcount, const GLint* basevertex, Bool feedDrawID,
+                                    Bool feedBaseVertex) {
+            const SizeT indexSize = IndexTypeSize(type);
+            if (indexSize == 0) return false;
+
+            SizeT total = 0;
+            for (GLsizei i = 0; i < drawcount; ++i) {
+                if (count[i] > 0) total += static_cast<SizeT>(count[i]);
+            }
+            if (total == 0) return true;
+            if (total > kMaxFlattenedIndices) return false;
+
+            const auto& indexBuffer = BoundIndexBuffer();
+            const Uint8* indexBufferBytes = nullptr;
+            SizeT indexBufferSize = 0;
+            if (indexBuffer) {
+                // The shadow is the source of truth for CPU reads, but a persistent map or
+                // a shader write may have moved past it since the last sync.
+                indexBuffer->SyncPersistentMappedRange();
+                indexBuffer->SyncGpuWrites();
+                indexBufferBytes = indexBuffer->MappedData();
+                indexBufferSize = indexBuffer->GetSize();
+            }
+
+            const Bool restartActive = RestartActive();
+            const Uint32 restartSentinel = RestartSentinelFor(type);
+            g_indexStaging.resize(total);
+            SizeT cursor = 0;
+            for (GLsizei i = 0; i < drawcount; ++i) {
+                if (count[i] <= 0) continue;
+                const SizeT subDrawCount = static_cast<SizeT>(count[i]);
+                const Uint8* source = ResolveSubDrawIndices(indexBuffer, indexBufferBytes, indexBufferSize, indices[i],
+                                                            subDrawCount, indexSize);
+                if (!source) {
+                    MGLOG_E_ONCE("DirectGLES multi-draw (drawelements tier): sub-draw %d reads outside the bound index "
+                            "buffer; skipping the batch",
+                            i);
+                    return false;
+                }
+                RebaseIndices(source, subDrawCount, indexSize, basevertex ? basevertex[i] : 0, restartActive,
+                              restartSentinel, g_indexStaging.data() + cursor);
+                cursor += subDrawCount;
+            }
+
+            SizeT indexBase = 0;
+            if (!UploadScratchRing(g_rebasedIndices, total * sizeof(Uint32), g_indexStaging.data(), indexBase)) {
+                return false;
+            }
+
+            const Uint previousIndexBinding = BoundIndexBufferId();
+            BufferImpl::BindBufferId(GL_ELEMENT_ARRAY_BUFFER, g_rebasedIndices.id);
+            cursor = 0;
+            for (GLsizei i = 0; i < drawcount; ++i) {
+                if (count[i] <= 0) continue;
+                if (feedDrawID) SetCurrentDrawID(static_cast<Uint32>(i));
+                // The base vertex is folded into the rewritten index stream here, so the
+                // driver sees none - but gl_BaseVertex still has to report the value the
+                // application passed for this sub-draw.
+                if (feedBaseVertex) SetCurrentBaseVertex(basevertex ? basevertex[i] : 0);
+                g_GLESFuncs.glDrawElements(mode, count[i], GL_UNSIGNED_INT,
+                                           reinterpret_cast<const void*>(indexBase + cursor * sizeof(Uint32)));
+                cursor += static_cast<SizeT>(count[i]);
+            }
+            if (feedDrawID) SetCurrentDrawID(0);
+            if (feedBaseVertex) SetCurrentBaseVertex(0);
+            BufferImpl::BindBufferId(GL_ELEMENT_ARRAY_BUFFER, previousIndexBinding);
+            NoteTierExecuted(GLESMultiDrawMode::DrawElements);
+            return true;
+        }
+
+        // ---------------------------------------------------------------------------
+        // Tier: Compute - the whole batch flattened into one rebased index stream
+        // ---------------------------------------------------------------------------
+
+        // One index per invocation. The sub-draw an output slot belongs to is found by
+        // binary search over the inclusive prefix sums of the sub-draw counts, which is
+        // why the descriptors are sorted by construction. Sub-draws with a zero count
+        // repeat the previous prefix sum and are therefore skipped by the search.
+        //
+        // Three storage blocks, not the five the shape suggests: ES 3.1 only guarantees
+        // four per compute stage, so the per-sub-draw descriptors share one buffer.
+        constexpr const char* kFlattenComputeSource = R"(#version 310 es
+layout(local_size_x = 64) in;
+
+uniform uint uElementSize;
+uniform uint uDrawCount;
+uniform uint uTotalIndices;
+
+layout(std430, binding = 0) readonly buffer SourceIndices { uint sourceWords[]; };
+layout(std430, binding = 1) readonly buffer DrawInfo { uint drawInfo[]; };
+layout(std430, binding = 2) writeonly buffer FlatIndices { uint flatIndices[]; };
+
+uint ReadSourceIndex(uint element) {
+    if (uElementSize == 4u) {
+        return sourceWords[element];
+    }
+    if (uElementSize == 2u) {
+        uint word = sourceWords[element >> 1u];
+        return (word >> ((element & 1u) * 16u)) & 0xFFFFu;
+    }
+    uint word = sourceWords[element >> 2u];
+    return (word >> ((element & 3u) * 8u)) & 0xFFu;
+}
+
+void main() {
+    uint outIndex = gl_GlobalInvocationID.x;
+    if (outIndex >= uTotalIndices) {
+        return;
+    }
+
+    uint low = 0u;
+    uint high = uDrawCount - 1u;
+    while (low < high) {
+        uint mid = low + (high - low) / 2u;
+        if (drawInfo[mid * 3u + 2u] > outIndex) {
+            high = mid;
+        } else {
+            low = mid + 1u;
+        }
+    }
+
+    uint localIndex = outIndex - (low == 0u ? 0u : drawInfo[(low - 1u) * 3u + 2u]);
+    // Unsigned wraparound is the defined behaviour for a negative base vertex. No
+    // restart sentinel handling: the tier declines outright while restart is enabled.
+    flatIndices[outIndex] = ReadSourceIndex(localIndex + drawInfo[low * 3u]) + drawInfo[low * 3u + 1u];
+}
+)";
+
+        struct FlattenedStream {
+            Uint bufferId = 0;
+            SizeT indexCount = 0;
+        };
+
+        Bool EnsureComputeProgram() {
+            if (g_computeProgram != 0) return true;
+            if (g_computeProgramFailed) return false;
+            g_computeProgramFailed = true; // cleared again only on a complete success
+
+            const GLuint shader = g_GLESFuncs.glCreateShader(GL_COMPUTE_SHADER);
+            if (shader == 0) {
+                MGLOG_E_ONCE("DirectGLES multi-draw (compute tier): glCreateShader(GL_COMPUTE_SHADER) failed");
+                return false;
+            }
+            const char* source = kFlattenComputeSource;
+            g_GLESFuncs.glShaderSource(shader, 1, &source, nullptr);
+            g_GLESFuncs.glCompileShader(shader);
+            GLint status = GL_FALSE;
+            g_GLESFuncs.glGetShaderiv(shader, GL_COMPILE_STATUS, &status);
+            if (status != GL_TRUE) {
+                char log[1024] = {};
+                g_GLESFuncs.glGetShaderInfoLog(shader, sizeof(log) - 1, nullptr, log);
+                MGLOG_E_ONCE("DirectGLES multi-draw (compute tier): index-flattening shader failed to compile: %s", log);
+                g_GLESFuncs.glDeleteShader(shader);
+                return false;
+            }
+
+            const GLuint program = g_GLESFuncs.glCreateProgram();
+            if (program == 0) {
+                MGLOG_E_ONCE("DirectGLES multi-draw (compute tier): glCreateProgram failed");
+                g_GLESFuncs.glDeleteShader(shader);
+                return false;
+            }
+            g_GLESFuncs.glAttachShader(program, shader);
+            g_GLESFuncs.glLinkProgram(program);
+            g_GLESFuncs.glDeleteShader(shader);
+            g_GLESFuncs.glGetProgramiv(program, GL_LINK_STATUS, &status);
+            if (status != GL_TRUE) {
+                char log[1024] = {};
+                g_GLESFuncs.glGetProgramInfoLog(program, sizeof(log) - 1, nullptr, log);
+                MGLOG_E_ONCE("DirectGLES multi-draw (compute tier): index-flattening program failed to link: %s", log);
+                g_GLESFuncs.glDeleteProgram(program);
+                return false;
+            }
+
+            g_computeProgram = program;
+            g_uElementSize = g_GLESFuncs.glGetUniformLocation(program, "uElementSize");
+            g_uDrawCount = g_GLESFuncs.glGetUniformLocation(program, "uDrawCount");
+            g_uTotalIndices = g_GLESFuncs.glGetUniformLocation(program, "uTotalIndices");
+            g_computeProgramFailed = false;
+            MGLOG_D("DirectGLES multi-draw: index-flattening compute program ready (id %u)", program);
+            return true;
+        }
+
+        // Builds the flattened stream, or leaves `out` empty when this batch's shape rules
+        // the tier out. Runs BEFORE PrepareForDraw - see the call site - so it may leave
+        // the compute program current and the first storage points unbound; the
+        // preparation that follows re-establishes both.
+        void FlattenWithCompute(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
+                                GLsizei drawcount, const GLint* basevertex, FlattenedStream& out) {
+            if (!SupportsTier(GLESMultiDrawMode::Compute)) return;
+            const SizeT indexSize = IndexTypeSize(type);
+            if (indexSize == 0) return;
+
+            // Merging sub-draws into a single draw only reproduces the original primitive
+            // stream for list-shaped modes: a strip, loop or fan would gain primitives
+            // spanning the seam between two sub-draws.
+            const Uint32 primitiveSize = ConcatenablePrimitiveSize(mode);
+            if (primitiveSize == 0) return;
+
+            // Primitive restart defeats the whole-multiple-of-a-primitive argument below,
+            // even for a list mode. A restart ends the current primitive, so a sub-draw of
+            // six GL_TRIANGLES indices with a restart after the third emits ONE triangle
+            // and drops the two leftover vertices - and once concatenated those leftovers
+            // find a third vertex in the next sub-draw and become a triangle that GL never
+            // draws. Splicing separator sentinels into the flattened stream could fix it,
+            // at the cost of a per-sub-draw offset the prefix-sum layout does not carry;
+            // declining is the honest trade for a tier that is already opt-in.
+            if (RestartActive()) return;
+
+            // The shader reads the source indices as a storage buffer, so there has to be
+            // a real buffer to read - a client-memory index array has none.
+            const auto& indexBuffer = BoundIndexBuffer();
+            if (!indexBuffer) return;
+
+            // A dispatch inside an open capture span is not legal, and the span would also
+            // observe one merged draw rather than the batch it asked for.
+            if (XfbImpl::IsCaptureSpanOpen()) return;
+
+            auto* sourceResource = BufferImpl::EnsureBufferResource(indexBuffer);
+            if (!sourceResource || sourceResource->id == 0) return;
+            const SizeT sourceSize = indexBuffer->GetSize();
+            // std430 addresses the source as uint[]; a tail shorter than a word is not
+            // reachable, so a narrow index type needs a word-multiple buffer.
+            if (indexSize < 4 && (sourceSize % 4) != 0) return;
+
+            g_drawInfoStaging.resize(3 * static_cast<SizeT>(drawcount));
+            SizeT total = 0;
+            for (GLsizei i = 0; i < drawcount; ++i) {
+                const SizeT subDrawCount = count[i] > 0 ? static_cast<SizeT>(count[i]) : 0;
+                // GL drops a trailing partial primitive per sub-draw; concatenation would
+                // instead splice it onto the next sub-draw's first vertices.
+                if (subDrawCount % primitiveSize != 0) return;
+                const SizeT byteOffset = reinterpret_cast<SizeT>(indices[i]);
+                if (byteOffset % indexSize != 0) return;
+                if (subDrawCount != 0) {
+                    const SizeT byteEnd = byteOffset + subDrawCount * indexSize;
+                    if (byteEnd > sourceSize || byteEnd < byteOffset) return;
+                }
+                total += subDrawCount;
+                if (total > kMaxComputeFlattenedIndices) return;
+                const SizeT slot = 3 * static_cast<SizeT>(i);
+                g_drawInfoStaging[slot] = static_cast<Uint32>(byteOffset / indexSize);
+                g_drawInfoStaging[slot + 1] = static_cast<Uint32>(basevertex ? basevertex[i] : 0);
+                g_drawInfoStaging[slot + 2] = static_cast<Uint32>(total);
+            }
+            if (total == 0) return; // nothing to draw; the ordinary tiers no-op just as well
+
+            if (!EnsureComputeProgram()) return;
+            if (!UploadScratch(g_drawInfo, g_drawInfoStaging.size() * sizeof(Uint32), g_drawInfoStaging.data())) {
+                return;
+            }
+            if (!UploadScratch(g_flattenedIndices, total * sizeof(Uint32), nullptr)) return;
+
+            BufferImpl::BindBufferBaseCached(GL_SHADER_STORAGE_BUFFER, 0, sourceResource->id);
+            BufferImpl::BindBufferBaseCached(GL_SHADER_STORAGE_BUFFER, 1, g_drawInfo.id);
+            BufferImpl::BindBufferBaseCached(GL_SHADER_STORAGE_BUFFER, 2, g_flattenedIndices.id);
+
+            g_GLESFuncs.glUseProgram(g_computeProgram);
+            PrgramImpl::g_lastUsedBackendProgramId = g_computeProgram;
+            if (g_uElementSize >= 0) g_GLESFuncs.glUniform1ui(g_uElementSize, static_cast<GLuint>(indexSize));
+            if (g_uDrawCount >= 0) g_GLESFuncs.glUniform1ui(g_uDrawCount, static_cast<GLuint>(drawcount));
+            if (g_uTotalIndices >= 0) g_GLESFuncs.glUniform1ui(g_uTotalIndices, static_cast<GLuint>(total));
+
+            g_GLESFuncs.glDispatchCompute(
+                static_cast<GLuint>((total + kComputeWorkGroupSize - 1) / kComputeWorkGroupSize), 1, 1);
+            g_GLESFuncs.glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT | GL_ELEMENT_ARRAY_BARRIER_BIT);
+
+            // Hand the storage points back to their GL default. PrepareForDraw re-syncs
+            // only the points the app has actually touched, so leaving a scratch buffer on
+            // an untouched point would keep it visible to the next shader that declares one.
+            for (Uint point = 0; point < 3; ++point) {
+                BufferImpl::BindBufferBaseCached(GL_SHADER_STORAGE_BUFFER, point, 0);
+            }
+
+            NoteTierExecuted(GLESMultiDrawMode::Compute);
+            out.bufferId = g_flattenedIndices.id;
+            out.indexCount = total;
+        }
+    } // namespace
+
+    // -------------------------------------------------------------------------------
+    // Public surface
+    // -------------------------------------------------------------------------------
+
+    Bool IsTierSupported(const MG_External::GLESCapabilities& caps, const MG_External::GLESFunctionsTable& funcs,
+                         GLESMultiDrawMode tier) {
+        const Bool esAtLeast31 =
+            caps.GLESVersion.Major > 3 || (caps.GLESVersion.Major == 3 && caps.GLESVersion.Minor >= 1);
+        switch (tier) {
+        case GLESMultiDrawMode::Ext:
+            return caps.SupportsMultiDrawElementsBaseVertex;
+        case GLESMultiDrawMode::MultiIndirect:
+            return caps.SupportsMultiDrawIndirect && esAtLeast31 && funcs.glDrawElementsIndirect != nullptr;
+        case GLESMultiDrawMode::Indirect:
+            return esAtLeast31 && funcs.glDrawElementsIndirect != nullptr;
+        case GLESMultiDrawMode::BaseVertex:
+            return caps.SupportsDrawElementsBaseVertex;
+        case GLESMultiDrawMode::DrawElements:
+            // Plain glDrawElements over a rewritten index stream: ES 2 core, so this is
+            // the floor every other tier can fall back to.
+            return true;
+        case GLESMultiDrawMode::Compute:
+            // Three storage blocks, which is inside the four ES 3.1 guarantees per stage.
+            return caps.SupportsComputeShader && caps.MaxComputeShaderStorageBlocks >= 3 &&
+                   funcs.glBindBufferBase != nullptr;
+        case GLESMultiDrawMode::Auto:
+            break;
+        }
+        return false;
+    }
+
+    GLESMultiDrawMode ResolveTier(const MG_External::GLESCapabilities& caps,
+                                  const MG_External::GLESFunctionsTable& funcs, GLESMultiDrawMode requested,
+                                  String* explanation) {
+        const auto bestAuto = [&]() {
+            for (const GLESMultiDrawMode tier : kAutoLadder) {
+                if (IsTierSupported(caps, funcs, tier)) return tier;
+            }
+            return GLESMultiDrawMode::DrawElements;
+        };
+
+        GLESMultiDrawMode resolved = GLESMultiDrawMode::DrawElements;
+        String line;
+        if (requested == GLESMultiDrawMode::Auto) {
+            resolved = bestAuto();
+            line = String("auto -> ") + TierName(resolved);
+        } else if (IsTierSupported(caps, funcs, requested)) {
+            resolved = requested;
+            line = String("MOBILEGL_ESPRYT_MULTIDRAW_MODE=") + TierName(requested) + " -> " + TierName(resolved);
+        } else {
+            resolved = bestAuto();
+            line = String("MOBILEGL_ESPRYT_MULTIDRAW_MODE=") + TierName(requested) +
+                   " requested but unsupported by this driver -> " + TierName(resolved);
+        }
+
+        if (explanation) {
+            String supported;
+            for (const GLESMultiDrawMode tier : kAutoLadder) {
+                if (!IsTierSupported(caps, funcs, tier)) continue;
+                if (!supported.empty()) supported += ", ";
+                supported += TierName(tier);
+            }
+            if (IsTierSupported(caps, funcs, GLESMultiDrawMode::Compute)) {
+                supported += supported.empty() ? "compute (opt-in)" : ", compute (opt-in)";
+            }
+            *explanation = line + " (driver supports: " + supported + ")";
+        }
+        return resolved;
+    }
+
+    const char* TierName(GLESMultiDrawMode tier) {
+        switch (tier) {
+        case GLESMultiDrawMode::Auto: return "auto";
+        case GLESMultiDrawMode::Ext: return "ext";
+        case GLESMultiDrawMode::MultiIndirect: return "multiindirect";
+        case GLESMultiDrawMode::Indirect: return "indirect";
+        case GLESMultiDrawMode::BaseVertex: return "basevertex";
+        case GLESMultiDrawMode::DrawElements: return "drawelements";
+        case GLESMultiDrawMode::Compute: return "compute";
+        }
+        return "unknown";
+    }
+
+    GLESMultiDrawMode ResolvedTier() {
+        ResolveTierOnce();
+        return g_resolvedTier;
+    }
+
+    String DescribeTierResolution() {
+        ResolveTierOnce();
+        return g_tierResolution;
+    }
+
+    void OnBackendContextDestroyed() {
+        g_indirectCommands = {};
+        g_rebasedIndices = {};
+        g_drawInfo = {};
+        g_flattenedIndices = {};
+        g_computeProgram = 0;
+        g_computeProgramFailed = false;
+        g_uElementSize = -1;
+        g_uDrawCount = -1;
+        g_uTotalIndices = -1;
+    }
+
+    void DrawElementsBatch(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
+                           GLsizei drawcount, const GLint* basevertex) {
+        if (drawcount <= 0 || !count || !indices) return;
+        // State-independent and possibly throwing, so it runs before any GL work.
+        CheckPrimitiveRestartSupported(type);
+
+        const Bool hasIndexBuffer = BoundIndexBuffer() != nullptr;
+
+        // The compute tier dispatches BEFORE the draw state is established: doing it
+        // afterwards would mean unpicking the program, SSBO and index bindings
+        // PrepareForDraw just made, and a dispatch inside an open transform feedback
+        // span is not legal at all. On success it hands back a flattened index stream.
+        // A batch whose sub-draws carry their own base vertices cannot be flattened either
+        // when the program reads gl_BaseVertex: one draw call leaves one uniform value.
+        // Asked conservatively because this decision precedes PrepareForDraw - see
+        // CurrentProgramMayNeedPerSubDrawBuiltins. Flattening is the irreversible half:
+        // once the batch is one draw the values are gone, whereas declining to flatten only
+        // costs the unrolled tier.
+        FlattenedStream flattened;
+        if (ResolvedTier() == GLESMultiDrawMode::Compute &&
+            !CurrentProgramMayNeedPerSubDrawBuiltins(basevertex != nullptr)) {
+            FlattenWithCompute(mode, count, type, indices, drawcount, basevertex, flattened);
+        }
+
+        PrepareForDraw(DrawSyncBit::IndexBuffer);
+
+        if (flattened.indexCount != 0) {
+            const Uint previousIndexBinding = BoundIndexBufferId();
+            BufferImpl::BindBufferId(GL_ELEMENT_ARRAY_BUFFER, flattened.bufferId);
+            g_GLESFuncs.glDrawElements(mode, static_cast<GLsizei>(flattened.indexCount), GL_UNSIGNED_INT, nullptr);
+            BufferImpl::BindBufferId(GL_ELEMENT_ARRAY_BUFFER, previousIndexBinding);
+            return;
+        }
+
+        // Now that PrepareForDraw has synced the program, both questions have real answers;
+        // the tier choice and the per-sub-draw feeds use those, not the guess above.
+        const Bool feedDrawID = CurrentProgramReadsDrawID();
+        const Bool feedBaseVertex = basevertex != nullptr && CurrentProgramReadsBaseVertex();
+        const GLESMultiDrawMode tier = ResolveTierForBatch(feedDrawID, feedBaseVertex, hasIndexBuffer);
+
+        Bool drawn = false;
+        switch (tier) {
+        case GLESMultiDrawMode::Ext:
+            drawn = RunExt(mode, count, type, indices, drawcount, basevertex);
+            break;
+        case GLESMultiDrawMode::MultiIndirect:
+            drawn = RunIndirect(mode, count, type, indices, drawcount, basevertex, /*batched=*/true, feedDrawID,
+                                feedBaseVertex);
+            break;
+        case GLESMultiDrawMode::Indirect:
+            drawn = RunIndirect(mode, count, type, indices, drawcount, basevertex, /*batched=*/false, feedDrawID,
+                                feedBaseVertex);
+            break;
+        case GLESMultiDrawMode::BaseVertex:
+            drawn = RunBaseVertexLoop(mode, count, type, indices, drawcount, basevertex, feedDrawID, feedBaseVertex);
+            break;
+        case GLESMultiDrawMode::DrawElements:
+            drawn = RunRebasedDrawElements(mode, count, type, indices, drawcount, basevertex, feedDrawID,
+                                           feedBaseVertex);
+            break;
+        case GLESMultiDrawMode::Compute:
+            // Its pre-pass ran above; reaching here means it declined this batch's shape.
+            break;
+        case GLESMultiDrawMode::Auto:
+            break; // resolution never yields Auto
+        }
+
+        // Every tier above may decline a batch whose shape it cannot express. The two
+        // below are the floor: a base-vertex replay where the driver has one, and the
+        // rewritten index stream where it does not. Both are safe for any batch these
+        // entry points can receive.
+        if (!drawn) {
+            drawn = RunBaseVertexLoop(mode, count, type, indices, drawcount, basevertex, feedDrawID, feedBaseVertex);
+        }
+        if (!drawn) {
+            drawn = RunRebasedDrawElements(mode, count, type, indices, drawcount, basevertex, feedDrawID,
+                                           feedBaseVertex);
+        }
+        if (!drawn) {
+            MGLOG_E_ONCE("DirectGLES multi-draw: no usable tier for a %d sub-draw batch (mode 0x%x, type 0x%x); "
+                    "the batch was dropped",
+                    drawcount, mode, type);
+        }
+    }
+} // namespace MobileGL::MG_Backend::DirectGLES::MultiDrawImpl

MobileGL/MG_Backend/DirectGLES/MultiDraw.h

@@ -0,0 +1,64 @@
+// MobileGL - MobileGL/MG_Backend/DirectGLES/MultiDraw.h
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#pragma once
+#include <Includes.h>
+#include <Config.h>
+#include "DirectGLES.h"
+
+// Emulation of the desktop glMultiDrawElements / glMultiDrawElementsBaseVertex entry
+// points on OpenGL ES, which has neither in core.
+//
+// Every strategy below is an emulation; they differ only in which driver capability
+// they lean on and in how many driver entries a batch of N sub-draws costs. The design
+// follows MobileGlues (MobileGL-Dev/MobileGlues, gl/multidraw.cpp) tier for tier, plus
+// the native GL_EXT_multi_draw_arrays interaction that MobileGL already had:
+//
+//   Ext           one glMultiDrawElementsBaseVertexEXT           1 driver entry
+//   MultiIndirect one glMultiDrawElementsIndirectEXT             1 driver entry + 1 upload
+//   Indirect      N x glDrawElementsIndirect                     N + 1 upload
+//   BaseVertex    N x glDrawElementsBaseVertex                   N
+//   DrawElements  N x glDrawElements over CPU-rebased indices    N + 1 upload
+//   Compute       1 x glDrawElements over a GPU-flattened,       1 dispatch + 1 entry
+//                 rebased index stream
+//
+// Which one runs is resolved once per ES context from the driver's capabilities,
+// capped by MOBILEGL_ESPRYT_MULTIDRAW_MODE, and can additionally be demoted per batch
+// when the batch's own shape rules a tier out (see ResolveTierForBatch in the .cpp).
+namespace MobileGL::MG_Backend::DirectGLES::MultiDrawImpl {
+    // The tier this ES context resolved to, computed on first use and stable after.
+    MG_Config::GLESMultiDrawMode ResolvedTier();
+    // "multiindirect", "compute", ... - stable identifiers, also used by the POST row.
+    const char* TierName(MG_Config::GLESMultiDrawMode tier);
+    // One line naming the resolved tier, the tiers the driver can support, and the env
+    // clamp if one applied. For DriverPost and the startup log.
+    String DescribeTierResolution();
+
+    // The resolution itself, as a pure function of a capability set: the backend feeds
+    // it the live ES context's capabilities, DriverPost feeds it the ones it probed
+    // standalone, and both therefore report the same tier. `explanation`, when non-null,
+    // receives the "requested -> resolved (driver supports: ...)" line.
+    MG_Config::GLESMultiDrawMode ResolveTier(const MG_External::GLESCapabilities& caps,
+                                             const MG_External::GLESFunctionsTable& funcs,
+                                             MG_Config::GLESMultiDrawMode requested, String* explanation);
+    // Whether one tier is runnable on the given capability set, for per-row POST output.
+    Bool IsTierSupported(const MG_External::GLESCapabilities& caps, const MG_External::GLESFunctionsTable& funcs,
+                         MG_Config::GLESMultiDrawMode tier);
+
+    // Runs `drawcount` indexed sub-draws as one glMultiDrawElements(BaseVertex) call
+    // would. `basevertex` is null for the plain glMultiDrawElements entry point (every
+    // base vertex is 0). Owns the whole draw, preparation included: callers must not
+    // have run PrepareForDraw, because the compute tier has to dispatch before the
+    // draw state is established.
+    void DrawElementsBatch(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
+                           GLsizei drawcount, const GLint* basevertex);
+
+    // The ES context is gone: every scratch buffer and the compute program belonged to
+    // it, so drop the names without deleting them (the dead context reclaims them).
+    void OnBackendContextDestroyed();
+} // namespace MobileGL::MG_Backend::DirectGLES::MultiDrawImpl

MobileGL/MG_Backend/DirectGLES/Utils.cpp

@@ -21,6 +21,7 @@
 #include <MG_Util/Math/HalfFloat.h>
 #include <MG_Util/Math/SmallFloat.h>
 
+#include <algorithm>
 #include <cmath>
 #include <cctype>
 #include <cstring>
@@ -61,29 +62,6 @@ namespace MobileGL::MG_Backend::DirectGLES {
                 requestedInternalFormat, GetDriverPixelFormatNormalizeOptions() | extraOptions);
         }
 
-        // Multisample textures can only ever be rendered into, never uploaded to, so a fallback
-        // format for them has to stay colour-renderable - a three-channel float fallback is a legal
-        // ES texture format but not a legal multisample storage format. Widening to four channels
-        // is safe here precisely because there is no transfer path that would have to expand
-        // three-channel client data, and the alpha the draw writes for a three-channel source is
-        // already the 1.0 the frontend format implies.
-        Bool TargetRequiresRenderableFormat(SizeT targetIndex) {
-            return targetIndex == static_cast<SizeT>(TextureTarget::Texture2DMultisample) ||
-                   targetIndex == static_cast<SizeT>(TextureTarget::Texture2DMultisampleArray);
-        }
-
-        Flags<PixelFormatNormalizeOptionBit> GetRenderTargetNormalizeOptions(SizeT targetIndex) {
-            Flags<PixelFormatNormalizeOptionBit> options;
-            if (!TargetRequiresRenderableFormat(targetIndex)) {
-                return options;
-            }
-            options |= PixelFormatNormalizeOptionBit::NoThreeChannelRenderTarget;
-            if (!g_GLESCapabilities.SupportsRenderSnorm || !g_GLESCapabilities.SupportsNorm16Texture) {
-                options |= PixelFormatNormalizeOptionBit::NoSnorm16RenderTarget;
-            }
-            return options;
-        }
-
         Bool HasCachedFormatCapability(TextureInternalFormat internalFormat,
                                        SizeT targetIndex,
                                        Bool caveat,
@@ -141,14 +119,61 @@ namespace MobileGL::MG_Backend::DirectGLES {
             const GLenum requestedInternalFormat = MG_Util::ConvertTextureInternalFormatToGLEnum(internalFormat);
             Flags<PixelFormatNormalizeOptionBit> options;
             if (!pActiveBackendObject || ShouldUseCaveatFormat(internalFormat, targetIndex)) {
-                options = GetRuntimeFallbackNormalizeOptions(requestedInternalFormat,
+                options = GetRuntimeFallbackNormalizeOptions(
-                                                             GetRenderTargetNormalizeOptions(targetIndex));
+                    requestedInternalFormat,
+                    TextureImpl::GetRenderTargetNormalizeOptions(g_GLESCapabilities, targetIndex));
             }
             NormalizePixelFormat(requestedInternalFormat, options, outInternalFormat, outFormat, outType);
         }
     } // namespace
 
     namespace TextureImpl {
+        // Every image that can back a colour attachment needs a colour-renderable storage format,
+        // and ES has no renderable three-channel format at all: a three-channel float fallback is
+        // a legal ES texture but neither legal multisample storage nor a legal attachment, so
+        // GL_RGB8_SNORM / GL_RGB16F / ... have to be widened to four channels for any of them.
+        // This used to cover the multisample pair alone, on the grounds that only those can never
+        // be uploaded to; the transfer paths now expand three-channel client data themselves
+        // (Managers.cpp PrepareFallbackUpload) and hide the added alpha again on sample and
+        // readback, so the same substitution is available everywhere.
+        //
+        // The widening only ever *happens* where the driver refuses the native form (see
+        // PopulateFormatCapabilitiesImpl: outside multisample storage it rides the driver branch,
+        // behind the native probe), so a driver that does render to a three-channel image keeps
+        // allocating it byte for byte.
+        //
+        // Do NOT read that as "nothing changes off-device". Measured on Mesa 26.1.6 llvmpipe
+        // (the headless CI driver), an ES 3.2 GL_TEXTURE_2D colour attachment is COMPLETE for
+        // GL_RGB8 and GL_RGB16F but INCOMPLETE_ATTACHMENT for GL_RGB8_SNORM, GL_SRGB8 and every
+        // RGB integer format, and UNSUPPORTED for GL_RGB32F. Those eight formats therefore DO
+        // take the widened path on llvmpipe, which is where the retrace fixtures and the glcts
+        // green suites run - the substitution is driver-conditional, not desktop-exempt.
+        //
+        // A buffer texture is the one image that can never be an attachment; its storage is the
+        // buffer object's, and widening it would misdescribe the application's data.
+        Bool TargetRequiresRenderableFormat(SizeT targetIndex) {
+            if (targetIndex >= kFormatCapabilityTargetCount) {
+                return false;
+            }
+            if (targetIndex == kFormatCapabilityRenderbufferTargetIndex) {
+                return true;
+            }
+            return static_cast<TextureTarget>(targetIndex) != TextureTarget::TextureBuffer;
+        }
+
+        Flags<PixelFormatNormalizeOptionBit> GetRenderTargetNormalizeOptions(
+            const MG_External::GLESCapabilities& capabilities, SizeT targetIndex) {
+            Flags<PixelFormatNormalizeOptionBit> options;
+            if (!TargetRequiresRenderableFormat(targetIndex)) {
+                return options;
+            }
+            options |= PixelFormatNormalizeOptionBit::NoThreeChannelRenderTarget;
+            if (!capabilities.SupportsRenderSnorm || !capabilities.SupportsNorm16Texture) {
+                options |= PixelFormatNormalizeOptionBit::NoSnorm16RenderTarget;
+            }
+            return options;
+        }
+
         void GenerateTextureFormatInfo(TextureInternalFormat internalFormat, GLenum* outInternalFormat,
                                        GLenum* outFormat, GLenum* outType, TextureTarget target) {
 #ifdef TRACY_ENABLE
@@ -178,9 +203,8 @@ namespace MobileGL::MG_Backend::DirectGLES {
             return ShouldUseCaveatFormat(internalFormat, GetRenderbufferFormatCapabilityTargetIndex());
         }
 
-        Bool BackendTextureFormatAddsAlpha(TextureInternalFormat internalFormat, TextureTarget target) {
+        namespace {
-            const SizeT targetIndex =
+            Bool BackendFormatAddsAlpha(TextureInternalFormat internalFormat, SizeT targetIndex) {
-                target == TextureTarget::Unknown ? kFormatCapabilityTargetCount : GetFormatCapabilityTargetIndex(target);
                 if (!TargetRequiresRenderableFormat(targetIndex)) {
                     return false;
                 }
@@ -188,11 +212,21 @@ namespace MobileGL::MG_Backend::DirectGLES {
                     return false;
                 }
                 const GLenum requestedInternalFormat = MG_Util::ConvertTextureInternalFormatToGLEnum(internalFormat);
-            const Flags<PixelFormatNormalizeOptionBit> options =
+                const Flags<PixelFormatNormalizeOptionBit> options = GetRuntimeFallbackNormalizeOptions(
-                GetRuntimeFallbackNormalizeOptions(requestedInternalFormat,
+                    requestedInternalFormat, GetRenderTargetNormalizeOptions(g_GLESCapabilities, targetIndex));
-                                                   GetRenderTargetNormalizeOptions(targetIndex));
                 return static_cast<Bool>(options & PixelFormatNormalizeOptionBit::NoThreeChannelRenderTarget);
             }
+        } // namespace
+
+        Bool BackendTextureFormatAddsAlpha(TextureInternalFormat internalFormat, TextureTarget target) {
+            const SizeT targetIndex =
+                target == TextureTarget::Unknown ? kFormatCapabilityTargetCount : GetFormatCapabilityTargetIndex(target);
+            return BackendFormatAddsAlpha(internalFormat, targetIndex);
+        }
+
+        Bool BackendRenderbufferFormatAddsAlpha(TextureInternalFormat internalFormat) {
+            return BackendFormatAddsAlpha(internalFormat, GetRenderbufferFormatCapabilityTargetIndex());
+        }
     } // namespace TextureImpl
     namespace PrgramImpl {
         String ProcessOutColorLocations(const String& glslCode) {
@@ -383,10 +417,27 @@ namespace MobileGL::MG_Backend::DirectGLES {
                 result = std::regex_replace(result, pattern, "$1flat $2");
             };
 
+            // Every stage that has an integer interface at all, on BOTH sides. Interpolation is
+            // only ever consumed at a fragment input, so the qualifier is semantically inert on
+            // a tessellation or geometry interface - but an ES linker still compares the two
+            // sides of every interface and rejects a program whose producer says `flat` and
+            // whose consumer does not. Covering only the stages that "need" it left exactly two
+            // holes, and a program that used tessellation fell into both:
+            //   vertex `flat out uint` -> tess-control `in uint`   (producer flat, consumer not)
+            //   tess-eval `out uint`   -> geometry `flat in uint`  (consumer flat, producer not)
+            // Adreno answers "output ... interpolation mismatch with other stage" and the whole
+            // program fails to link, which is a draw that silently paints nothing.
+            //
+            // Adding rather than stripping, because a fragment input's `flat` is load-bearing
+            // (ESSL forbids an interpolated integer) and would have to be put back for the last
+            // stage before the fragment shader anyway - so "everything integer is flat" is the
+            // one rule that is consistent no matter which stages a program happens to have.
             switch (shaderType) {
             case GL_VERTEX_SHADER:
                 addFlatQualifier("out");
                 break;
+            case GL_TESS_CONTROL_SHADER:
+            case GL_TESS_EVALUATION_SHADER:
             case GL_GEOMETRY_SHADER:
                 addFlatQualifier("in");
                 addFlatQualifier("out");
@@ -401,6 +452,175 @@ namespace MobileGL::MG_Backend::DirectGLES {
             return result;
         }
 
+        String RetargetTextureBufferExtension(String glslCode,
+                                              MG_External::GLESCapabilities::TextureBufferTier tier) {
+#ifdef TRACY_ENABLE
+            ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+            // SPIRV-Cross hardcodes the EXT spelling: CompilerGLSL::type_to_glsl emits
+            // require_extension_internal("GL_EXT_texture_buffer") for any Dim=Buffer image
+            // whenever it targets ESSL below 320, with no OES alternative and no way to
+            // configure it. GL_OES_texture_buffer is functionally identical but is a separate
+            // directive, and `#extension <name> : require` on a name the driver does not
+            // advertise is a hard compile error - so on an OES-only driver the emitted shader
+            // fails to compile for the sake of one token.
+            //
+            // Line comments are excluded by the directive check below; a `#extension` line inside
+            // a /* */ block is not, and would be rewritten. That is harmless (it stays a comment)
+            // and is not worth a preprocessor-aware scan here.
+            //
+            // Deliberately a directive rewrite and nothing more. The alternative - teaching the
+            // SPIR-V to stop asking for the extension - is not available: the requirement is
+            // synthesized by SPIRV-Cross from the image type itself, not carried in the module,
+            // so there is nothing upstream to strip. Everything about the shader body that
+            // actually uses the buffer texture is identical between the two extensions.
+            using Tier = MG_External::GLESCapabilities::TextureBufferTier;
+            if (tier != Tier::ExtensionOES) {
+                return glslCode;
+            }
+            static constexpr const char* kExtName = "GL_EXT_texture_buffer";
+            static constexpr const char* kOesName = "GL_OES_texture_buffer";
+            constexpr SizeT kExtNameLength = 21; // strlen("GL_EXT_texture_buffer")
+            static_assert(sizeof("GL_EXT_texture_buffer") - 1 == kExtNameLength, "name length drifted");
+            static_assert(sizeof("GL_OES_texture_buffer") - 1 == kExtNameLength,
+                          "the two spellings must be the same length for the in-place replace");
+
+            // Only rewrite the name where it is the whole subject of an #extension directive.
+            // Two separate guards, both load-bearing:
+            //   * the directive check, so a line-comment mentioning the name is left alone;
+            //   * the identifier-boundary check, because GL_EXT_texture_buffer is a PREFIX of
+            //     GL_EXT_texture_buffer_object - a different, real extension that SPIRV-Cross
+            //     emits from the same `case DimBuffer:` on its legacy-desktop branch. Without
+            //     the boundary this pass would silently rewrite a request for that extension
+            //     into a request for a GL_OES_texture_buffer_object that does not exist.
+            const auto isIdentifierChar = [](char c) {
+                return std::isalnum(static_cast<unsigned char>(c)) != 0 || c == '_';
+            };
+            SizeT searchFrom = 0;
+            while (true) {
+                const SizeT hit = glslCode.find(kExtName, searchFrom);
+                if (hit == String::npos) {
+                    break;
+                }
+                searchFrom = hit + kExtNameLength;
+
+                // Identifier boundary on both sides, so the name is not a fragment of a longer one.
+                if (hit > 0 && isIdentifierChar(glslCode[hit - 1])) {
+                    continue;
+                }
+                if (hit + kExtNameLength < glslCode.size() && isIdentifierChar(glslCode[hit + kExtNameLength])) {
+                    continue;
+                }
+
+                // Walk back to the start of the line and require that it is an #extension
+                // directive, allowing whitespace between '#' and the keyword.
+                SizeT lineStart = glslCode.rfind('\n', hit);
+                lineStart = (lineStart == String::npos) ? 0 : lineStart + 1;
+                SizeT cursor = lineStart;
+                while (cursor < hit && std::isspace(static_cast<unsigned char>(glslCode[cursor]))) {
+                    ++cursor;
+                }
+                if (cursor >= hit || glslCode[cursor] != '#') {
+                    continue;
+                }
+                ++cursor;
+                while (cursor < hit && std::isspace(static_cast<unsigned char>(glslCode[cursor]))) {
+                    ++cursor;
+                }
+                if (glslCode.compare(cursor, 9, "extension") != 0) {
+                    continue;
+                }
+                glslCode.replace(hit, kExtNameLength, kOesName);
+            }
+            return glslCode;
+        }
+
+        String RequestExtendedImageFormats(String glslCode, Bool needed) {
+#ifdef TRACY_ENABLE
+            ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+            // GLSL ES core has thirteen image formats; GL has forty. SPIRV-Cross prints whatever
+            // format the OpTypeImage carries and asks for no extension for it, so an r8ui or
+            // rg16f image - declared as such, or baked from the bound one - reaches the driver as
+            // a format its core language does not know. GL_NV_image_formats is the only thing
+            // that adds them, and it has to be requested by name.
+            //
+            // The caller decides `needed`: it knows which formats are in play (from the uniform
+            // reflection and the image-unit bindings) and whether the driver advertises the
+            // extension at all - `#extension` on an unadvertised name is itself a hard error, so
+            // this must never be emitted speculatively.
+            static constexpr const char* kDirective = "#extension GL_NV_image_formats : require\n";
+            static constexpr const char* kExtName = "GL_NV_image_formats";
+            if (!needed || glslCode.find(kExtName) != String::npos) {
+                return glslCode;
+            }
+            // After the #version line, which must stay first. Everything else about the header is
+            // order-insensitive, and ForceSupporterOutput's scan for the LAST #extension
+            // directive still finds whichever one that is.
+            const SizeT versionPos = glslCode.find("#version");
+            if (versionPos == String::npos) {
+                return kDirective + glslCode;
+            }
+            const SizeT lineEnd = glslCode.find('\n', versionPos);
+            if (lineEnd == String::npos) {
+                return glslCode + "\n" + kDirective;
+            }
+            glslCode.insert(lineEnd + 1, kDirective);
+            return glslCode;
+        }
+
+        String BakeImageFormatQualifiers(String glslCode,
+                                         const UnorderedMap<String, String>& esslFormatByUniformName) {
+#ifdef TRACY_ENABLE
+            ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+            if (esslFormatByUniformName.empty() || glslCode.find("image") == String::npos) {
+                return glslCode;
+            }
+            // Same declaration shape RebindImageUniformsToFrontendUnits matches, and for the same
+            // reason: one line, one image uniform, the name in group 3.
+            static const std::regex imageDeclRegex(
+                R"((layout\s*\(([^)]*)\)\s*)?uniform\s+(?:(?:readonly|writeonly|coherent|volatile|restrict|highp|mediump|lowp)\s+)*[iu]?image[A-Za-z0-9]+\s+([A-Za-z_][A-Za-z0-9_]*)\s*(\[[^\]]*\])?\s*;)");
+            // Every image format spelling GLSL has, so a declaration that already carries one is
+            // recognised whatever it says - the caller's map is consulted only for declarations
+            // with NO format, never to override a written one.
+            static const std::regex existingFormatRegex(
+                R"(\b(rgba32f|rgba16f|rg32f|rg16f|r11f_g11f_b10f|r32f|r16f|rgba16|rgb10_a2|rg16|rg8|r16|r8|rgba16_snorm|rgba8_snorm|rg16_snorm|rg8_snorm|r16_snorm|r8_snorm|rgba32i|rgba16i|rgba8i|rg32i|rg16i|rg8i|r32i|r16i|r8i|rgba32ui|rgba16ui|rgba8ui|rgb10_a2ui|rg32ui|rg16ui|rg8ui|r32ui|r16ui|r8ui)\b)");
+
+            String result;
+            result.reserve(glslCode.size());
+            SizeT lineStart = 0;
+            while (lineStart <= glslCode.size()) {
+                const SizeT lineEnd = glslCode.find('\n', lineStart);
+                const Bool lastLine = lineEnd == String::npos;
+                String line = glslCode.substr(lineStart, lastLine ? String::npos : lineEnd - lineStart);
+
+                std::smatch match;
+                if (std::regex_search(line, match, imageDeclRegex)) {
+                    const String name = match[3].str();
+                    const auto formatIt = esslFormatByUniformName.find(name);
+                    const String layoutContents = match[2].matched ? match[2].str() : String();
+                    if (formatIt != esslFormatByUniformName.end() && !formatIt->second.empty() &&
+                        !std::regex_search(layoutContents, existingFormatRegex)) {
+                        if (match[1].matched) {
+                            const SizeT layoutOpen = line.find('(', match.position(1));
+                            line.insert(layoutOpen + 1, formatIt->second + ", ");
+                        } else {
+                            line.insert(match.position(0), "layout(" + formatIt->second + ") ");
+                        }
+                    }
+                }
+
+                result += line;
+                if (lastLine) {
+                    break;
+                }
+                result += '\n';
+                lineStart = lineEnd + 1;
+            }
+            return result;
+        }
+
         String RemoveLayoutBinding(const String& glslCode) {
 #ifdef TRACY_ENABLE
             ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
@@ -437,6 +657,352 @@ namespace MobileGL::MG_Backend::DirectGLES {
             return result;
         }
 
+        namespace {
+            Bool IsImagePassIdentifierChar(char c) {
+                return std::isalnum(static_cast<unsigned char>(c)) || c == '_';
+            }
+
+            // Occurrences of `identifier` in `code` that are whole identifiers, i.e. not the
+            // tail or head of a longer one. "goku" must not find "goku_hd" or "my_goku".
+            SizeT CountIdentifierOccurrences(const String& code, const String& identifier) {
+                if (identifier.empty()) return 0;
+                SizeT count = 0;
+                for (SizeT pos = code.find(identifier); pos != String::npos;
+                     pos = code.find(identifier, pos + 1)) {
+                    if (pos > 0 && IsImagePassIdentifierChar(code[pos - 1])) continue;
+                    const SizeT after = pos + identifier.size();
+                    if (after < code.size() && IsImagePassIdentifierChar(code[after])) continue;
+                    ++count;
+                }
+                return count;
+            }
+
+            Bool ContainsIdentifier(const String& code, const String& identifier) {
+                return CountIdentifierOccurrences(code, identifier) > 0;
+            }
+
+            // The image format layout qualifiers ESSL accepts (GLSL ES 3.20 4.4.7 table 4.6 -
+            // the ES-legal subset of what SPIRV-Cross's format_to_glsl can print). The
+            // readonly/writeonly rule only applies to a declaration that carries one of them.
+            Bool IsImageFormatQualifier(const String& token) {
+                static constexpr StringView FORMATS[] = {
+                    "rgba32f",      "rgba16f",      "rg32f",       "rg16f",        "r11f_g11f_b10f",
+                    "r32f",         "r16f",         "rgba16",      "rgb10_a2",     "rgba8",
+                    "rg16",         "rg8",          "r16",         "r8",           "rgba16_snorm",
+                    "rgba8_snorm",  "rg16_snorm",   "rg8_snorm",   "r16_snorm",    "r8_snorm",
+                    "rgba32i",      "rgba16i",      "rgba8i",      "rg32i",        "rg16i",
+                    "rg8i",         "r32i",         "r16i",        "r8i",          "rgba32ui",
+                    "rgba16ui",     "rgb10_a2ui",   "rgba8ui",     "rg32ui",       "rg16ui",
+                    "rg8ui",        "r32ui",        "r16ui",       "r8ui",
+                };
+                for (const StringView format : FORMATS) {
+                    if (token == format) return true;
+                }
+                return false;
+            }
+
+            // "Except for image variables qualified with the format qualifiers r32f, r32i, and
+            // r32ui, image variables must specify either memory qualifier readonly or the
+            // memory qualifier writeonly." (GLSL ES 3.20 4.10)
+            Bool IsMemoryQualifierExemptImageFormat(const String& token) {
+                return token == "r32f" || token == "r32i" || token == "r32ui";
+            }
+
+            // Comma-separated contents of a layout(...) list, each entry trimmed.
+            Vector<String> SplitLayoutQualifierList(const String& layout) {
+                Vector<String> tokens;
+                SizeT start = 0;
+                while (start <= layout.size()) {
+                    SizeT comma = layout.find(',', start);
+                    const Bool last = comma == String::npos;
+                    String token = layout.substr(start, last ? String::npos : comma - start);
+                    const SizeT first = token.find_first_not_of(" \t\r\n");
+                    if (first == String::npos) {
+                        token.clear();
+                    } else {
+                        token = token.substr(first, token.find_last_not_of(" \t\r\n") - first + 1);
+                    }
+                    if (!token.empty()) tokens.push_back(Move(token));
+                    if (last) break;
+                    start = comma + 1;
+                }
+                return tokens;
+            }
+
+            // Trims both ends and collapses every internal whitespace run to one space, so a
+            // qualifier list or array suffix can be spliced back into a rebuilt declaration
+            // whatever the original spacing was.
+            String NormalizeDeclarationSpacing(const String& text) {
+                String out;
+                out.reserve(text.size());
+                Bool pendingSpace = false;
+                for (const char c : text) {
+                    if (std::isspace(static_cast<unsigned char>(c))) {
+                        pendingSpace = !out.empty();
+                        continue;
+                    }
+                    if (pendingSpace) out += ' ';
+                    pendingSpace = false;
+                    out += c;
+                }
+                return out;
+            }
+
+            // How an image builtin touches the image it is handed.
+            enum class ImageBuiltinAccess { None, Load, Store, Unknown };
+
+            ImageBuiltinAccess ClassifyImageBuiltin(const String& name) {
+                if (name == "imageStore") return ImageBuiltinAccess::Store;
+                if (name == "imageLoad") return ImageBuiltinAccess::Load;
+                // imageAtomic* both reads and writes, but ES only defines the atomics on
+                // r32i/r32ui/r32f images - exactly the formats the rule above exempts - so this
+                // pass has already skipped any declaration they can legally appear on. Load is
+                // enough to keep the classification total without ever being acted upon.
+                if (name.compare(0, 11, "imageAtomic") == 0) return ImageBuiltinAccess::Load;
+                if (name == "imageSize" || name == "imageSamples") return ImageBuiltinAccess::None;
+                // Some other identifier that starts with "image" and is being called: not a
+                // shape this pass can reason about, so it poisons the declaration instead of
+                // being guessed at.
+                return ImageBuiltinAccess::Unknown;
+            }
+
+            struct ImageUniformDecl {
+                String name;
+                String writeName;   // the writeonly half's name, when split
+                String layout;      // raw contents of layout(...)
+                String qualifiers;  // memory/precision qualifiers, normalized, no trailing space
+                String type;        // image2D, uimage2DArray, ...
+                String arraySuffix; // "" or "[7]"
+                SizeT declStart = 0;
+                SizeT declLength = 0;
+                SizeT referenceCount = 0; // uses this pass recognized and accounted for
+                Bool loaded = false;
+                Bool stored = false;
+                Bool unknownUse = false;
+                Bool split = false;
+            };
+
+            // A rebuilt declaration. Keeps SPIRV-Cross's own word order (`uniform readonly
+            // highp image2D`) so the image-rebinding regex in Managers.cpp still matches what
+            // comes out of here, whichever order the two passes end up running in.
+            String BuildImageDeclaration(const ImageUniformDecl& decl, const char* memoryQualifier,
+                                         const String& variableName) {
+                String out = "layout(" + decl.layout + ") uniform ";
+                out += memoryQualifier;
+                out += ' ';
+                if (!decl.qualifiers.empty()) {
+                    out += decl.qualifiers;
+                    out += ' ';
+                }
+                out += decl.type;
+                out += ' ';
+                out += variableName;
+                out += decl.arraySuffix;
+                out += ';';
+                return out;
+            }
+
+            // A name for the writeonly half that no identifier in the shader (and no other
+            // half already minted) can collide with.
+            String MakeImageWriteAliasName(const String& name, const String& source,
+                                           const Vector<String>& taken) {
+                String candidate = String(IMAGE_WRITE_ALIAS_PREFIX) + name;
+                // "__" anywhere in an identifier is reserved (GLSL ES 3.20 3.7), which a name
+                // that already starts with '_' would otherwise produce.
+                for (SizeT doubled = candidate.find("__"); doubled != String::npos;
+                     doubled = candidate.find("__", doubled)) {
+                    candidate.erase(doubled, 1);
+                }
+                auto isTaken = [&](const String& identifier) {
+                    if (ContainsIdentifier(source, identifier)) return true;
+                    for (const auto& other : taken) {
+                        if (other == identifier) return true;
+                    }
+                    return false;
+                };
+                while (isTaken(candidate)) candidate += 'X';
+                return candidate;
+            }
+
+            struct ImageSourceEdit {
+                SizeT start;
+                SizeT length;
+                String text;
+            };
+        } // namespace
+
+        String SplitReadWriteImageUniforms(const String& glslCode) {
+#ifdef TRACY_ENABLE
+            ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+            if (glslCode.find("image") == String::npos) {
+                return glslCode;
+            }
+
+            // layout(...) uniform <memory/precision qualifiers> <image type> <name>[array];
+            // The qualifier alternation is order-free even though SPIRV-Cross emits a fixed
+            // order (to_qualifiers_glsl: storage, then coherent/restrict/readonly/writeonly,
+            // then precision), and the array group is repeated so a hypothetical multi-
+            // dimensional image array survives the round trip intact.
+            static const std::regex imageDeclRegex(
+                R"(layout\s*\(([^)]*)\)\s*uniform\s+)"
+                R"(((?:(?:readonly|writeonly|coherent|volatile|restrict|highp|mediump|lowp)\s+)*))"
+                R"(([iu]?image[A-Za-z0-9_]*)\s+([A-Za-z_][A-Za-z0-9_]*)\s*((?:\[[^\]]*\]\s*)*);)");
+
+            Vector<ImageUniformDecl> decls;
+            for (std::sregex_iterator it(glslCode.begin(), glslCode.end(), imageDeclRegex), last; it != last; ++it) {
+                const std::smatch& match = *it;
+                const String qualifiers = match[2].str();
+                // Already legal: SPIRV-Cross decided one way, leave it alone.
+                if (ContainsIdentifier(qualifiers, "readonly") || ContainsIdentifier(qualifiers, "writeonly")) {
+                    continue;
+                }
+
+                Bool hasFormat = false;
+                Bool exemptFormat = false;
+                for (const String& token : SplitLayoutQualifierList(match[1].str())) {
+                    if (!IsImageFormatQualifier(token)) continue;
+                    hasFormat = true;
+                    exemptFormat = IsMemoryQualifierExemptImageFormat(token);
+                }
+                // No format qualifier at all is a different (and, in ES, unconditionally
+                // illegal) shape that GL_EXT_shader_image_load_formatted would be needed for;
+                // SPIRV-Cross refuses to emit it for an ES target, so nothing to do here.
+                if (!hasFormat || exemptFormat) continue;
+
+                ImageUniformDecl decl;
+                decl.layout = match[1].str();
+                decl.qualifiers = NormalizeDeclarationSpacing(qualifiers);
+                decl.type = match[3].str();
+                decl.name = match[4].str();
+                decl.arraySuffix = NormalizeDeclarationSpacing(match[5].str());
+                decl.declStart = static_cast<SizeT>(match.position(0));
+                decl.declLength = match[0].str().size();
+                decls.push_back(Move(decl));
+            }
+            if (decls.empty()) {
+                return glslCode;
+            }
+
+            auto findDecl = [&decls](const String& name) -> SizeT {
+                for (SizeT i = 0; i < decls.size(); ++i) {
+                    if (decls[i].name == name) return i;
+                }
+                return decls.size();
+            };
+
+            // Walk every `image*(` call and attribute its first argument to a declaration.
+            struct StoreSite {
+                SizeT declIndex;
+                SizeT start;
+                SizeT length;
+            };
+            Vector<StoreSite> storeSites;
+            for (SizeT pos = glslCode.find("image"); pos != String::npos; pos = glslCode.find("image", pos + 1)) {
+                if (pos > 0 && IsImagePassIdentifierChar(glslCode[pos - 1])) continue; // uimage2D, myimageFoo
+                SizeT tokenEnd = pos;
+                while (tokenEnd < glslCode.size() && IsImagePassIdentifierChar(glslCode[tokenEnd])) ++tokenEnd;
+                const String builtin = glslCode.substr(pos, tokenEnd - pos);
+
+                const SizeT openParen = glslCode.find_first_not_of(" \t\r\n", tokenEnd);
+                if (openParen == String::npos || glslCode[openParen] != '(') continue; // a type, not a call
+
+                const SizeT argStart = glslCode.find_first_not_of(" \t\r\n", openParen + 1);
+                if (argStart == String::npos) continue;
+                if (!std::isalpha(static_cast<unsigned char>(glslCode[argStart])) && glslCode[argStart] != '_') {
+                    continue; // an expression, not a bare variable - it names no image of ours
+                }
+                SizeT argEnd = argStart;
+                while (argEnd < glslCode.size() && IsImagePassIdentifierChar(glslCode[argEnd])) ++argEnd;
+
+                const SizeT declIndex = findDecl(glslCode.substr(argStart, argEnd - argStart));
+                if (declIndex == decls.size()) continue;
+                ImageUniformDecl& decl = decls[declIndex];
+                ++decl.referenceCount;
+
+                // The operand has to be the bare variable, optionally subscripted. Anything
+                // else (a member access, a call result) is a shape this pass cannot rewrite.
+                SizeT after = glslCode.find_first_not_of(" \t\r\n", argEnd);
+                if (after != String::npos && glslCode[after] == '[') {
+                    Int depth = 0;
+                    SizeT scan = after;
+                    for (; scan < glslCode.size(); ++scan) {
+                        if (glslCode[scan] == '[') ++depth;
+                        else if (glslCode[scan] == ']' && --depth == 0) break;
+                    }
+                    after = scan >= glslCode.size() ? String::npos
+                                                    : glslCode.find_first_not_of(" \t\r\n", scan + 1);
+                }
+                const char nextChar = after == String::npos ? '\0' : glslCode[after];
+                if (nextChar != ',' && nextChar != ')') {
+                    decl.unknownUse = true;
+                    continue;
+                }
+
+                switch (ClassifyImageBuiltin(builtin)) {
+                case ImageBuiltinAccess::Load:
+                    decl.loaded = true;
+                    break;
+                case ImageBuiltinAccess::Store:
+                    decl.stored = true;
+                    storeSites.push_back({declIndex, argStart, argEnd - argStart});
+                    break;
+                case ImageBuiltinAccess::None:
+                    break;
+                default:
+                    decl.unknownUse = true;
+                    break;
+                }
+            }
+
+            // Every mention of the name has to be one this pass saw, or the split would leave
+            // a store pointing at the readonly half. One occurrence is the declaration itself.
+            for (auto& decl : decls) {
+                if (CountIdentifierOccurrences(glslCode, decl.name) != decl.referenceCount + 1) {
+                    decl.unknownUse = true;
+                }
+            }
+
+            Vector<ImageSourceEdit> edits;
+            Vector<String> takenAliases;
+            for (auto& decl : decls) {
+                if (decl.unknownUse) continue; // leave it exactly as it was; no guessing
+                if (decl.loaded && decl.stored) {
+                    decl.writeName = MakeImageWriteAliasName(decl.name, glslCode, takenAliases);
+                    takenAliases.push_back(decl.writeName);
+                    decl.split = true;
+                    edits.push_back({decl.declStart, decl.declLength,
+                                     BuildImageDeclaration(decl, "readonly", decl.name) + "\n" +
+                                         BuildImageDeclaration(decl, "writeonly", decl.writeName)});
+                } else if (decl.stored) {
+                    edits.push_back({decl.declStart, decl.declLength,
+                                     BuildImageDeclaration(decl, "writeonly", decl.name)});
+                } else {
+                    // Loaded only, or only ever handed to imageSize (or unused): readonly is
+                    // the qualifier that keeps every one of those legal.
+                    edits.push_back({decl.declStart, decl.declLength,
+                                     BuildImageDeclaration(decl, "readonly", decl.name)});
+                }
+            }
+            for (const StoreSite& site : storeSites) {
+                const ImageUniformDecl& decl = decls[site.declIndex];
+                if (!decl.split) continue;
+                edits.push_back({site.start, site.length, decl.writeName});
+            }
+            if (edits.empty()) {
+                return glslCode;
+            }
+
+            // Back to front, so an earlier edit's offsets stay valid.
+            std::sort(edits.begin(), edits.end(),
+                      [](const ImageSourceEdit& a, const ImageSourceEdit& b) { return a.start > b.start; });
+            String result = glslCode;
+            for (const ImageSourceEdit& edit : edits) {
+                result.replace(edit.start, edit.length, edit.text);
+            }
+            return result;
+        }
+
         namespace {
             // How a lookup carries its level of detail, and how many arguments it takes
             // before the optional bias.
@@ -494,7 +1060,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
             }
         } // namespace
 
-        String EmulateTextureLodBias(const String& glslCode) {
+        String EmulateTextureLodBias(const String& glslCode, Bool avoidExplicitLodBias) {
 #ifdef TRACY_ENABLE
             ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
 #endif
@@ -555,6 +1121,11 @@ namespace MobileGL::MG_Backend::DirectGLES {
                 if (samplerIt == samplerNames.end()) continue;
 
                 const String& biasName = samplerIt->second;
+                if (form->explicitLodArg >= 0 && avoidExplicitLodBias) {
+                    // The lookup already names its level; leaving it alone keeps a constant
+                    // LOD constant. Costs the bias on explicit-LOD lookups only.
+                    continue;
+                }
                 if (form->explicitLodArg >= 0) {
                     // Explicit LOD: the bias adds to it, as Vulkan does for
                     // OpImageSampleExplicitLod and as the CTS reference expects.
@@ -605,7 +1176,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
             ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
 #endif
             for (GLenum err = g_GLESFuncs.glGetError(); err != GL_NO_ERROR; err = g_GLESFuncs.glGetError()) {
-                MGLOG_E("-> GLES Error: %s", MG_Util::ConvertGLEnumToString(err).c_str());
+                MGLOG_D("-> GLES Error: %s", MG_Util::ConvertGLEnumToString(err).c_str());
             }
         }
 
@@ -1025,26 +1596,18 @@ namespace MobileGL::MG_Backend::DirectGLES {
             return (rowBytes + align - 1) / align * align;
         }
 
-    // Repacks wide RGBA(_INTEGER) rows into the client's (format, type) layout, honoring the
+        // Walks the client-side destination the PACK parameters describe and hands each row to
-        // client-side PACK parameters and the bound pixel-pack buffer. `wide` holds
+        // `fillRow(slice, row, dstRow)`, which writes width * dstPixelBytes bytes of finished client
-        // `sliceHeight * sliceCount` rows of `width` texels (slice-major, tightly stacked),
+        // texels. Shared by the converting and the raw-word stores so both address the destination -
-        // 4 components x GetReadbackComponentSize(wideType) bytes each.
+        // and feed the bound pixel-pack buffer - identically.
         // applyPackImageParams: GL_PACK_IMAGE_HEIGHT / GL_PACK_SKIP_IMAGES apply only to GetTexImage
         // of 3D/array images; ReadPixels and 2D GetTexImage ignore them (GL 3.3 sections 4.3.1, 6.1.4).
         // Per the GL addressing rules, slice k row j lands at
         // SKIP_IMAGES*imageStride + SKIP_ROWS*rowStride + SKIP_PIXELS*pixelBytes
         //   + k*imageStride + j*rowStride, with imageStride = max(IMAGE_HEIGHT, sliceHeight)*rowStride.
-        Bool StoreWideRowsToClient(const Uint8* wide, GLenum wideType, GLsizei width, GLsizei sliceHeight,
+        template <typename FillRow>
-                                   GLsizei sliceCount, const ReadbackChannelMapping& mapping, GLenum type,
+        static Bool StoreClientRows(SizeT dstPixelBytes, SizeT swapGroupSize, GLsizei width, GLsizei sliceHeight,
-                                   void* pixels, Bool applyPackImageParams) {
+                                    GLsizei sliceCount, void* pixels, Bool applyPackImageParams, FillRow&& fillRow) {
-        const SizeT dstPixelBytes = GetReadbackDstPixelSize(mapping, type);
-        if (dstPixelBytes == 0) {
-            return false;
-        }
-        PackedReadbackLayout packedLayout{};
-        const Bool isPackedType = GetPackedReadbackLayout(type, packedLayout);
-        const SizeT dstComponentSize = GetReadbackComponentSize(type);
-
             const auto& pixelPackBufferObject =
                 MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::PixelPack).GetBoundObject();
 
@@ -1071,29 +1634,20 @@ namespace MobileGL::MG_Backend::DirectGLES {
                                         static_cast<SizeT>(sliceCount - 1) * dstImageStride +
                                         static_cast<SizeT>(sliceHeight - 1) * dstRowStride + dstRowBytes;
                 if (requiredSize > pixelPackBufferObject->GetSize()) {
-                MGLOG_E("Readback conversion: pixel pack buffer is too small");
+                    MGLOG_E_ONCE("Readback conversion: pixel pack buffer is too small");
                     return true;
                 }
             }
 
-        const SizeT srcComponentSize = GetReadbackComponentSize(wideType);
-        const SizeT srcPixelBytes = 4 * srcComponentSize;
             Vector<Uint8> convertedRow(dstRowBytes);
 
             for (GLsizei slice = 0; slice < sliceCount; ++slice) {
                 for (GLsizei row = 0; row < sliceHeight; ++row) {
-                const SizeT flatRow = static_cast<SizeT>(slice) * static_cast<SizeT>(sliceHeight) +
+                    fillRow(slice, row, convertedRow.data());
-                                   static_cast<SizeT>(row);
-                const Uint8* srcRow = wide + flatRow * static_cast<SizeT>(width) * srcPixelBytes;
-                ConvertWideReadbackRow(srcRow, convertedRow.data(), static_cast<SizeT>(width), wideType,
-                                                  mapping, type);
 
-                if (packParams.SwapBytes) {
+                    if (packParams.SwapBytes && swapGroupSize > 1) {
-                    const SizeT groupSize = isPackedType ? packedLayout.byteSize : dstComponentSize;
+                        for (SizeT offset = 0; offset + swapGroupSize <= dstRowBytes; offset += swapGroupSize) {
-                    if (groupSize > 1) {
+                            std::reverse(convertedRow.data() + offset, convertedRow.data() + offset + swapGroupSize);
-                        for (SizeT offset = 0; offset + groupSize <= dstRowBytes; offset += groupSize) {
-                            std::reverse(convertedRow.data() + offset, convertedRow.data() + offset + groupSize);
-                        }
                         }
                     }
 
@@ -1107,7 +1661,59 @@ namespace MobileGL::MG_Backend::DirectGLES {
                     }
                 }
             }
+            if (pixelPackBufferObject) {
+                // WritebackFromBackend bumps change serials with no backend op; re-open
+                // the buffer draw-clean memos (once for the whole row loop).
+                BufferImpl::BumpBufferMutationEpoch();
+            }
             return true;
         }
+
+        // Repacks wide RGBA(_INTEGER) rows into the client's (format, type) layout, honoring the
+        // client-side PACK parameters and the bound pixel-pack buffer. `wide` holds
+        // `sliceHeight * sliceCount` rows of `width` texels (slice-major, tightly stacked),
+        // 4 components x GetReadbackComponentSize(wideType) bytes each.
+        Bool StoreWideRowsToClient(const Uint8* wide, GLenum wideType, GLsizei width, GLsizei sliceHeight,
+                                   GLsizei sliceCount, const ReadbackChannelMapping& mapping, GLenum type,
+                                   void* pixels, Bool applyPackImageParams) {
+            const SizeT dstPixelBytes = GetReadbackDstPixelSize(mapping, type);
+            if (dstPixelBytes == 0) {
+                return false;
+            }
+            PackedReadbackLayout packedLayout{};
+            const Bool isPackedType = GetPackedReadbackLayout(type, packedLayout);
+            const SizeT swapGroupSize = isPackedType ? packedLayout.byteSize : GetReadbackComponentSize(type);
+            const SizeT srcPixelBytes = 4 * GetReadbackComponentSize(wideType);
+
+            return StoreClientRows(dstPixelBytes, swapGroupSize, width, sliceHeight, sliceCount, pixels,
+                                   applyPackImageParams,
+                                   [&](GLsizei slice, GLsizei row, Uint8* dstRow) {
+                                       const SizeT flatRow = static_cast<SizeT>(slice) *
+                                                                 static_cast<SizeT>(sliceHeight) +
+                                                             static_cast<SizeT>(row);
+                                       const Uint8* srcRow =
+                                           wide + flatRow * static_cast<SizeT>(width) * srcPixelBytes;
+                                       ConvertWideReadbackRow(srcRow, dstRow, static_cast<SizeT>(width), wideType,
+                                                              mapping, type);
+                                   });
+        }
+
+        Bool StorePackedWordsToClient(const Uint8* srcWords, GLsizei width, GLsizei sliceHeight, GLsizei sliceCount,
+                                      GLenum type, void* pixels, Bool applyPackImageParams) {
+            PackedReadbackLayout packedLayout{};
+            if (!GetPackedReadbackLayout(type, packedLayout) || packedLayout.byteSize != 4) {
+                return false;
+            }
+            const SizeT srcRowBytes = static_cast<SizeT>(width) * 4;
+
+            return StoreClientRows(4, packedLayout.byteSize, width, sliceHeight, sliceCount, pixels,
+                                   applyPackImageParams,
+                                   [&](GLsizei slice, GLsizei row, Uint8* dstRow) {
+                                       const SizeT flatRow = static_cast<SizeT>(slice) *
+                                                                 static_cast<SizeT>(sliceHeight) +
+                                                             static_cast<SizeT>(row);
+                                       Memcpy(dstRow, srcWords + flatRow * srcRowBytes, srcRowBytes);
+                                   });
+        }
     } // namespace ReadbackImpl
 } // namespace MobileGL::MG_Backend::DirectGLES

MobileGL/MG_Backend/DirectGLES/Utils.h

@@ -9,6 +9,8 @@
 #pragma once
 #include <Includes.h>
 #include <MG_State/GLState/Core.h>
+#include <MG_Util/BackendLoaders/OpenGL/Loader.h>
+#include <MG_Util/Texture/TextureFormatProcessor.h>
 
 namespace MobileGL::MG_Backend::DirectGLES {
     namespace DebugImpl {
@@ -34,6 +36,16 @@ namespace MobileGL::MG_Backend::DirectGLES {
     } // namespace VertexArrayImpl
 
     namespace TextureImpl {
+        // Whether images on this format-capability target can back a colour attachment, and so
+        // need a colour-renderable storage format even when the frontend asked for a
+        // three-channel one ES never renders to. Shared by the capability probe (which passes the
+        // capabilities it has just queried, before the globals are published) and by the
+        // allocation path (which reads the active backend's), so the format the cache was probed
+        // with is always the format the image is created with.
+        Bool TargetRequiresRenderableFormat(SizeT targetIndex);
+        Flags<PixelFormatNormalizeOptionBit> GetRenderTargetNormalizeOptions(
+            const MG_External::GLESCapabilities& capabilities, SizeT targetIndex);
+
         void GenerateTextureFormatInfo(TextureInternalFormat internalFormat, GLenum* outInternalFormat,
                                        GLenum* outFormat, GLenum* outType,
                                        TextureTarget target = TextureTarget::Unknown);
@@ -41,10 +53,12 @@ namespace MobileGL::MG_Backend::DirectGLES {
                                             GLenum* outFormat, GLenum* outType);
         Bool ShouldUseCaveatTextureFormat(TextureInternalFormat internalFormat, TextureTarget target);
 
-        // True when the format the texture is actually created with has an alpha channel the
+        // True when the format the image is actually created with has an alpha channel the
-        // frontend format does not (the three-channel multisample widening). GL reads such a
+        // frontend format does not (the three-channel colour-renderable widening). GL reads such
-        // channel back as 1.0, so any swizzle source of ALPHA has to be answered with ONE.
+        // a channel back as 1.0, so any swizzle source of ALPHA has to be answered with ONE and
+        // any readback of the image has to overwrite the alpha the draw happened to leave there.
         Bool BackendTextureFormatAddsAlpha(TextureInternalFormat internalFormat, TextureTarget target);
+        Bool BackendRenderbufferFormatAddsAlpha(TextureInternalFormat internalFormat);
         Bool ShouldUseCaveatRenderbufferFormat(TextureInternalFormat internalFormat);
     } // namespace TextureImpl
 
@@ -101,6 +115,16 @@ namespace MobileGL::MG_Backend::DirectGLES {
         Bool StoreWideRowsToClient(const Uint8* wide, GLenum wideType, GLsizei width, GLsizei sliceHeight,
                                    GLsizei sliceCount, const ReadbackChannelMapping& mapping, GLenum type,
                                    void* pixels, Bool applyPackImageParams);
+
+        // Stores packed 32-bit source words verbatim, with the same destination addressing, PACK
+        // parameters and pixel-pack-buffer handling as StoreWideRowsToClient. For the sources whose
+        // storage word already IS the client word (MG_Util::IsRawPackedPixelTransfer): routing those
+        // through the wide float intermediate re-encodes them, and the RGB9_E5 encoder canonicalizes
+        // the shared exponent, so glGetTexImage would answer with different bits than were stored.
+        // `srcWords` holds sliceHeight * sliceCount tightly stacked rows of `width` 32-bit words.
+        // False when `type` is not a 4-byte packed type.
+        Bool StorePackedWordsToClient(const Uint8* srcWords, GLsizei width, GLsizei sliceHeight, GLsizei sliceCount,
+                                      GLenum type, void* pixels, Bool applyPackImageParams);
     } // namespace ReadbackImpl
 
     namespace PrgramImpl {
@@ -116,7 +140,68 @@ namespace MobileGL::MG_Backend::DirectGLES {
         // drawBufferCount <= 1, i.e. for everything but a framebuffer that actually
         // enables several draw buffers, so the ordinary single-target shader is untouched.
         String BroadcastLegacyFragColor(String glslCode, GLenum shaderType, Uint drawBufferCount);
+        // SPIRV-Cross emits `#extension GL_EXT_texture_buffer : require` for every buffer-texture
+        // sampler when it targets ESSL below 320, and offers no way to ask for the OES spelling.
+        // On a driver that advertises only GL_OES_texture_buffer that directive is a compile
+        // error, so the name is retargeted in the emitted source. A no-op on every other tier:
+        // ES 3.2 needs no directive at all and an EXT driver already has the right one.
+        String RetargetTextureBufferExtension(String glslCode,
+                                              MG_External::GLESCapabilities::TextureBufferTier tier);
+        // Adds `#extension GL_NV_image_formats : require` when the shader carries an image
+        // format qualifier GLSL ES has no core spelling for. SPIRV-Cross prints the format and
+        // asks for nothing, so the request has to be made here. `needed` is the caller's answer,
+        // because only it knows which formats are in play AND whether the driver advertises the
+        // extension - requesting an unadvertised extension is itself a compile error, so this is
+        // never emitted speculatively. A no-op when not needed or already present.
+        String RequestExtendedImageFormats(String glslCode, Bool needed);
+        // Writes a format layout qualifier into the image declarations named in
+        // `esslFormatByUniformName` that still have none. The completion half of the image-format
+        // bake, and ONLY that: the SPIR-V pass (BakeImageFormatsPass) is what normally puts the
+        // format in, but SPIRV-Cross throws rather than printing the formats it calls
+        // desktop-only when it targets ESSL - r8ui among them, which is what the stencil half of
+        // KHR-GL4x.packed_depth_stencil.stencil_texturing binds - and a throw loses the whole
+        // stage. So those formats stay out of the module and are spelled here instead, on the
+        // emitted text, where nothing can refuse them.
+        //
+        // Declarations that already carry a format are left exactly as they are, whoever wrote
+        // it. Must run before RemoveLayoutBinding, which is where an image's layout qualifier
+        // stops being safe to edit by hand.
+        String BakeImageFormatQualifiers(String glslCode, const UnorderedMap<String, String>& esslFormatByUniformName);
         String RemoveLayoutBinding(const String& glslCode);
+        // Prefix of the writeonly half a read+write image uniform is split into (see
+        // SplitReadWriteImageUniforms); the suffix is the image's own name.
+        constexpr const char* IMAGE_WRITE_ALIAS_PREFIX = "mg_imageWrite_";
+        // ESSL refuses an image variable that carries a format qualifier other than r32f /
+        // r32i / r32ui unless it also carries `readonly` or `writeonly` (GLSL ES 3.10 4.9 /
+        // 3.20 4.10; glslang enforces it verbatim in ParseHelper.cpp's layoutObjectCheck).
+        // SPIRV-Cross emits NEITHER for an image the shader both reads and writes: it
+        // speculatively decorates every storage image NonWritable+NonReadable
+        // (fixup_image_load_store_access), then OpImageRead clears NonReadable and
+        // OpImageWrite clears NonWritable, and to_qualifiers_glsl only prints `readonly`
+        // from NonWritable and `writeonly` from NonReadable. Desktop GLSL is happy with the
+        // bare declaration, so the frontend raises no error and the illegal ESSL only shows
+        // up as a device compile failure - and then as a silently no-op draw.
+        //
+        // Restores a legal declaration:
+        //  * loaded only            -> add `readonly`
+        //  * stored only            -> add `writeonly`
+        //  * both                   -> emit TWO declarations on the same binding and of the
+        //                              same type, `readonly <name>` and `writeonly
+        //                              <IMAGE_WRITE_ALIAS_PREFIX><name>`, and point every
+        //                              imageStore at the second one. Several image variables
+        //                              may share an image unit as long as they have the same
+        //                              type and format, which is exactly what the pair is.
+        //
+        // Budget note: the split DOUBLES the image-uniform count of the stage it fires in, so
+        // a driver advertising a tight GL_MAX_{FRAGMENT,VERTEX,...}_IMAGE_UNIFORMS can turn a
+        // shader that used to compile into a link failure. ES only guarantees 4 fragment image
+        // uniforms, so a shader with more than half the limit in read+write images is the case
+        // to watch.
+        //
+        // Runs on the transpiled ESSL, so it must see the bindings the frontend units were
+        // already rewritten to and must run before those bindings are stripped - see the call
+        // site in Managers.cpp.
+        String SplitReadWriteImageUniforms(const String& glslCode);
         // Prefix of the per-sampler float uniform that carries GL_TEXTURE_LOD_BIAS into
         // the shader (see EmulateTextureLodBias); the suffix is the sampler's own name.
         constexpr const char* LOD_BIAS_UNIFORM_PREFIX = "mg_lodBias_";
@@ -128,7 +213,12 @@ namespace MobileGL::MG_Backend::DirectGLES {
         // the bound texture's (or sampler object's) value into it; a shader whose samplers
         // all have a zero bias is therefore unaffected. Returns the source unchanged when
         // there is nothing to rewrite.
-        String EmulateTextureLodBias(const String& glslCode);
+        //
+        // avoidExplicitLodBias leaves lookups that already carry an explicit LOD untouched,
+        // so their constant level stays constant; only the implicit-LOD forms take the bias.
+        // Off by default and only ever set on ANGLE + llvmpipe, where injecting the uniform
+        // into a constant LOD crashes the driver (MOBILEGL_AVOID_EXPLICIT_LOD_BIAS).
+        String EmulateTextureLodBias(const String& glslCode, Bool avoidExplicitLodBias = false);
     } // namespace PrgramImpl
 
     namespace Utils {

MobileGL/MG_Backend/DirectVulkan/BackendObject_DirectVulkan.cpp

@@ -9,13 +9,16 @@
 #include "BackendObject_DirectVulkan.h"
 #include "MG_Backend/BackendObject.h"
 #include "DirectVulkan.h"
+#include "SubgroupSupportPolicy.h"
 #include "MG_State/GLState/FramebufferState/FramebufferObject.h"
+#include "MG_State/GLState/Core.h"
 #include "MG_State/GLState/TextureState/TextureState.h"
 #include "MG_Util/Classifiers/TextureEnumClassifier.h"
 #include "MG_Util/Converters/MGToGL/TextureEnumConverter.h"
 #include "MG_Util/Converters/MGToStr/TextureEnumConverter.h"
 #include "MG_Util/Converters/MGToVk/TextureEnumConverter.h"
 #include "MG_Util/Texture/TextureFormatProcessor.h"
+#include "MG_Util/Async/ShaderCompilePool.h"
 
 #include <Config.h>
 #include <cmath>
@@ -382,6 +385,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
         UpdateDynamicBackendParameters();
         UpdateAdvertisedExtensions();
+        if (MG_State::pGLContext) {
+            MG_State::pGLContext->InvalidateCompileEnv();
+        }
         PopulateFormatCapabilities(physicalDevice.handle, vkGetPhysicalDeviceFormatProperties, m_vulkanCaps,
                                    MutableFormatCapabilities());
         PrintFormatCapabilities(GetFormatCapabilities());
@@ -496,33 +502,75 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             .ExtraVendor = Nullopt,
             .RendererGLInfo = {.TargetGLVersion = {4, 0, 0},
                                .TargetGLSLVersion = {4, 6, 0},
-                               // Baseline advertisement (no shader subgroup, no timer queries); a
+                               // Baseline advertisement (no runtime-gated capabilities); a live
-                               // live backend reconciles its copy in UpdateAdvertisedExtensions.
+                               // backend reconciles its copy in UpdateAdvertisedExtensions.
-                               .Extensions = BuildAdvertisedExtensions(false, false, false),
+                               .Extensions = BuildAdvertisedExtensions(false, false, false, false),
                                .IsCompatibilityProfile = false},
             .StaticBackendCapability = {.AllowVSOnlyPrograms = false}};
         return rendererInfo;
     }
 
     Vector<GLExtension> BuildAdvertisedExtensions(Bool shaderSubgroupSupported, Bool timerQueriesSupported,
-                                                  Bool anisotropicFilteringSupported) {
+                                                  Bool anisotropicFilteringSupported,
+                                                  Bool nonZeroIndirectBaseInstanceSupported) {
         Vector<GLExtension> extensions = {
             V_OpenGL30, V_OpenGL31, V_OpenGL32, V_OpenGL33, V_OpenGL40, E_GL_ARB_draw_buffers_blend,
             E_GL_ARB_compute_shader, E_GL_ARB_shader_storage_buffer_object, E_GL_ARB_shader_image_load_store,
-            E_GL_ARB_program_interface_query, E_GL_ARB_framebuffer_object, E_GL_ARB_multi_draw_indirect,
+            E_GL_ARB_clear_buffer_object, E_GL_ARB_program_interface_query, E_GL_ARB_framebuffer_object, E_GL_ARB_draw_indirect,
+            E_GL_ARB_multi_draw_indirect,
             E_GL_ARB_indirect_parameters, E_GL_EXT_framebuffer_object, E_GL_ARB_depth_texture, E_GL_ARB_buffer_storage,
             E_GL_ARB_texture_storage, E_GL_ARB_texture_storage_multisample, E_GL_ARB_texture_multisample,
             E_GL_ARB_clear_texture, E_GL_ARB_direct_state_access, E_GL_ARB_shader_draw_parameters,
             E_GL_ARB_gpu_shader_int64, E_GL_KHR_debug, E_GL_ARB_gpu_shader5, E_GL_ARB_multi_bind,
             E_GL_ARB_shading_language_420pack, E_GL_ARB_vertex_attrib_binding, E_GL_ARB_shader_image_size,
             E_GL_ARB_explicit_attrib_location,
+            // Core since GL 3.1 and implemented for every version advertised here. The string
+            // matters because applications gate the ENTRY POINTS on it rather than on the
+            // version: a caller that finds the extension missing never resolves
+            // glGetUniformBlockIndex / glUniformBlockBinding, and one that then uses uniform
+            // blocks anyway calls through a null pointer.
+            E_GL_ARB_uniform_buffer_object,
+            // Sampling the stencil aspect through DEPTH_STENCIL_TEXTURE_MODE. Core from 4.3,
+            // so on a 4.0 context the string is the only way to reach it.
+            E_GL_ARB_stencil_texturing,
             // Advertised with GL_NUM_PROGRAM_BINARY_FORMATS = 0, which the
             // extension explicitly permits. It is also the only thing that
             // exposes glProgramParameteri before GL 4.1.
             E_GL_ARB_get_program_binary};
+        // Vulkan's drawIndirectFirstInstance feature is optional. Direct base-instance calls work
+        // without it, but ARB_base_instance also promises non-zero firstInstance in GPU indirect
+        // commands; the renderer supplies true only when that word is legal and gl_InstanceID can
+        // be rebased to OpenGL's zero-based semantics.
+        if (nonZeroIndirectBaseInstanceSupported) {
+            extensions.push_back(E_GL_ARB_base_instance);
+        }
         if (shaderSubgroupSupported && !MG_Config::Features.DisableSubgroup) {
             extensions.push_back(E_GL_KHR_shader_subgroup);
         }
+        // GL_KHR_parallel_shader_compile is MobileGL's own capability, not the Vulkan
+        // device's: the compiler threads belong to MobileGL's shader pool and
+        // glCompileShader/glLinkProgram are serviced entirely inside the frontend, so there
+        // is no device feature to condition this on.
+        //
+        // Gated on the async flag deliberately, and this is the whole reason the gate
+        // exists. Advertising the string is the one part of asynchronous compilation that a
+        // recorded trace can never cover: Iris and Sodium change their SUBMISSION SCHEDULE
+        // the moment they see it - they enqueue whole pipeline batches and poll
+        // GL_COMPLETION_STATUS_KHR instead of compiling one program at a time - so
+        // MOBILEGL_ASYNC_SHADER_COMPILE=0 has to withdraw the application-visible behaviour
+        // change as well as the threading, or the kill switch would only be half a switch.
+        if (MG_Util::Async::AsyncShaderCompileEnabled()) {
+            extensions.push_back(E_GL_KHR_parallel_shader_compile);
+        }
+        // GL_ARB_gpu_shader_fp64 is opt-in (MOBILEGL_ADVERTISE_FP64). Every `double` in a
+        // shader compiles and runs already - it is narrowed to 32 bits before the module
+        // reaches this backend - so an application that simply uses doubles needs nothing
+        // advertised. What the extension additionally promises is 64-bit PRECISION, which no
+        // mobile GPU has and the narrowing cannot fake, so advertising it by default would
+        // make an application that checks the string take a path MobileGL cannot honour.
+        if (MG_Config::Features.AdvertiseFp64) {
+            extensions.push_back(E_GL_ARB_gpu_shader_fp64);
+        }
         // GL_ARB_timer_query gates MC's F3 GPU% (LWJGL checks the extension string);
         // only advertised when the device actually supports timestamp queries and the
         // MOBILEGL_DISABLE_TIMERQUERY escape hatch is off.
@@ -601,12 +649,6 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             funcsTable.GL.GetIntegeri_v = GetIntegeri_v;
             funcsTable.GL.GetInteger64i_v = GetInteger64i_v;
             funcsTable.GL.GetProgramiv = GetProgramiv;
-            funcsTable.GL.GetProgramInterfaceiv = GetProgramInterfaceiv;
-            funcsTable.GL.GetProgramResourceIndex = GetProgramResourceIndex;
-            funcsTable.GL.GetProgramResourceName = GetProgramResourceName;
-            funcsTable.GL.GetProgramResourceiv = GetProgramResourceiv;
-            funcsTable.GL.GetProgramResourceLocation = GetProgramResourceLocation;
-            funcsTable.GL.GetProgramResourceLocationIndex = GetProgramResourceLocationIndex;
             funcsTable.GL.ShaderStorageBlockBinding = ShaderStorageBlockBinding;
             funcsTable.GL.FenceSync = FenceSync;
             funcsTable.GL.ClientWaitSync = ClientWaitSync;
@@ -650,6 +692,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_vulkanCaps = capabilities;
         UpdateDynamicBackendParameters();
         UpdateAdvertisedExtensions();
+        if (MG_State::pGLContext) {
+            MG_State::pGLContext->InvalidateCompileEnv();
+        }
         MutableFormatCapabilities().Clear();
     }
 
@@ -660,9 +705,16 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // real device timestamp support. ApplyVulkanCapabilitiesForTesting may
         // run without a renderer; no timer query is advertised then. Rebuilding
         // the whole list keeps re-runs idempotent.
+        // The opt-in emulated compute path (SubgroupSupportPolicy.h) carries the
+        // extension by itself on devices with no native subgroup support at all; a
+        // device with native subgroups always advertises - and uses - those.
+        const Bool subgroupSupportAdvertised =
+            m_vulkanCaps.SupportsShaderSubgroup ||
+            ShouldEmulateSubgroups(m_vulkanCaps.SupportsShaderSubgroup);
         m_rendererInfo.RendererGLInfo.Extensions = BuildAdvertisedExtensions(
-            m_vulkanCaps.SupportsShaderSubgroup, pVulkanRenderer && pVulkanRenderer->IsTimerQuerySupported(),
+            subgroupSupportAdvertised, pVulkanRenderer && pVulkanRenderer->IsTimerQuerySupported(),
-            pVulkanRenderer && pVulkanRenderer->IsSamplerAnisotropySupported());
+            pVulkanRenderer && pVulkanRenderer->IsSamplerAnisotropySupported(),
+            pVulkanRenderer && pVulkanRenderer->IsNonZeroIndirectBaseInstanceSupported());
     }
 
     void BackendObject_DirectVulkan::UpdateDynamicBackendParameters() {
@@ -758,14 +810,48 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // the Uint32 attribute masks the draw path passes around are both bounded by MAX_VERTEX_ATTRIBS.
         m_dynamicParameters.MaxVertexAttribs = std::min(
             m_vulkanCaps.MaxVertexAttribs, static_cast<Int>(MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS));
-        m_dynamicParameters.MaxComputeShaderStorageBlocks = m_vulkanCaps.MaxComputeShaderStorageBlocks;
+        // Vulkan descriptor limits are not GL limits, and a GL application reads an advertised
-        m_dynamicParameters.MaxCombinedShaderStorageBlocks = m_vulkanCaps.MaxCombinedShaderStorageBlocks;
+        // limit as an amount it may actually USE. Adreno answers the per-stage/per-set descriptor
-        m_dynamicParameters.MaxComputeUniformBlocks = m_vulkanCaps.MaxComputeUniformBlocks;
+        // queries at descriptor-indexing scale - the same driver whose
+        // GL_MAX_SHADER_STORAGE_BLOCK_SIZE is clamped from 2147483647 further down - so
+        // KHR-GL44.multi_bind.dispatch_bind_buffers_base read GL_MAX_COMPUTE_UNIFORM_BLOCKS,
+        // created that many buffers and spliced that many UBO declarations into a single compute
+        // shader: ~14 s of allocation, then death on std::bad_alloc. Its sibling
+        // dispatch_bind_buffers_range hard-codes 4 buffers and passes, which is the clean
+        // discriminator. Every ceiling below is far above what any desktop driver advertises for
+        // these (84-96 for the binding families) and far below a descriptor-indexing count, so it
+        // can only lower a limit that was never usable in the first place. The zero floor is not
+        // decoration: a driver reporting UINT32_MAX used to arrive here as -1.
+        const auto clampLimit = [](const char* name, Int reported, Int ceiling) {
+            const Int clamped = std::min(std::max(reported, 0), ceiling);
+            if (clamped != reported) {
+                MGLOG_I("DirectVulkan: clamped %s from %d to %d", name, reported, clamped);
+            }
+            return clamped;
+        };
+        // GL 4.6 required minimums, for the record: MAX_COMPUTE_UNIFORM_BLOCKS 12,
+        // MAX_COMPUTE/COMBINED_SHADER_STORAGE_BLOCKS 8, MAX_SHADER_STORAGE_BUFFER_BINDINGS 8,
+        // MAX_UNIFORM_BUFFER_BINDINGS 84, MAX_TEXTURE_BUFFER_SIZE 65536.
+        constexpr Int kMaxAdvertisedBufferBlocks = 256;
+        constexpr Int kMaxAdvertisedTextureBufferSize = 1 << 27; // texels; what desktop GL reports
+        m_dynamicParameters.MaxComputeShaderStorageBlocks =
+            clampLimit("GL_MAX_COMPUTE_SHADER_STORAGE_BLOCKS", m_vulkanCaps.MaxComputeShaderStorageBlocks,
+                       kMaxAdvertisedBufferBlocks);
+        m_dynamicParameters.MaxCombinedShaderStorageBlocks =
+            clampLimit("GL_MAX_COMBINED_SHADER_STORAGE_BLOCKS", m_vulkanCaps.MaxCombinedShaderStorageBlocks,
+                       kMaxAdvertisedBufferBlocks);
+        m_dynamicParameters.MaxComputeUniformBlocks =
+            clampLimit("GL_MAX_COMPUTE_UNIFORM_BLOCKS", m_vulkanCaps.MaxComputeUniformBlocks,
+                       kMaxAdvertisedBufferBlocks);
         m_dynamicParameters.MaxComputeWorkGroupInvocations = m_vulkanCaps.MaxComputeWorkGroupInvocations;
-        m_dynamicParameters.MaxShaderStorageBufferBindings = m_vulkanCaps.MaxShaderStorageBufferBindings;
+        m_dynamicParameters.MaxShaderStorageBufferBindings =
-        m_dynamicParameters.MaxTextureBufferSize = m_vulkanCaps.MaxTextureBufferSize;
+            clampLimit("GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS", m_vulkanCaps.MaxShaderStorageBufferBindings,
+                       kMaxAdvertisedBufferBlocks);
+        m_dynamicParameters.MaxTextureBufferSize = clampLimit(
+            "GL_MAX_TEXTURE_BUFFER_SIZE", m_vulkanCaps.MaxTextureBufferSize, kMaxAdvertisedTextureBufferSize);
         m_dynamicParameters.TextureBufferOffsetAlignment = m_vulkanCaps.TextureBufferOffsetAlignment;
-        m_dynamicParameters.MaxUniformBufferBindings = m_vulkanCaps.MaxUniformBufferBindings;
+        m_dynamicParameters.MaxUniformBufferBindings = clampLimit(
+            "GL_MAX_UNIFORM_BUFFER_BINDINGS", m_vulkanCaps.MaxUniformBufferBindings, kMaxAdvertisedBufferBlocks);
         m_dynamicParameters.MaxUniformBlockSize = m_vulkanCaps.MaxUniformBlockSize;
         m_dynamicParameters.MaxImageUnits = std::max(std::min(m_vulkanCaps.MaxImageUnits, maxSupportedTextureUnits), 0);
         m_dynamicParameters.MaxCombinedImageUniforms = std::max(m_vulkanCaps.MaxCombinedImageUniforms, 0);
@@ -810,6 +896,50 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             }
         }
         m_dynamicParameters.SupportsWideLines = m_vulkanCaps.SupportsWideLines;
+        // A 2D or 2D multisample array texture is a VK_IMAGE_TYPE_2D image whose GL depth IS its
+        // arrayLayers, so a GL layer is a Vulkan array layer with nothing to translate.
+        // ResolveAttachmentBaseArrayLayer already passes the attachment's layer through. The other
+        // layered targets are declared separately as their own machinery lands.
+        {
+            using DynParams = MG_Backend::DynamicBackendParameters;
+            m_dynamicParameters.PerLayerFramebufferAttachmentTargets |=
+                DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::Texture2DArray) |
+                DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::Texture2DMultisampleArray);
+            // A cube map array is one 2D image with arrayLayers = 6 * cubeCount, so a GL layer is a
+            // Vulkan array layer here too - but the image cannot be created without imageCubeArray.
+            // A 3D texture's GL layer is a z slice, which only a 2D view over a 2D-array-compatible
+            // image can name. Optimistic: a format that refuses the flag is caught at image creation
+            // and declines the slice view there, which the clear path handles as a soft miss.
+            if (m_vulkanCaps.Supports2DArrayCompatible3DImages) {
+                m_dynamicParameters.PerLayerFramebufferAttachmentTargets |=
+                    DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::Texture3D);
+            }
+            if (m_vulkanCaps.SupportsImageCubeArray) {
+                m_dynamicParameters.PerLayerFramebufferAttachmentTargets |=
+                    DynParams::PerLayerFramebufferAttachmentBit(TextureTarget::TextureCubeMapArray);
+            }
+        }
+        // Never, on any device, and no longer for the reason it used to be. It used to track
+        // shaderFloat64 because a `dvec3` input needed the Float64 capability to exist in the
+        // module at all; a 64-bit vertex FETCH was already impossible (VK_FORMAT_R64*_SFLOAT is
+        // optional and lavapipe reports zero bufferFeatures for all four), so the attribute
+        // arrived as its 32-bit word pair and PackDoubleVertexInputsPass bitcast it back.
+        //
+        // The shader half of that is gone: every 64-bit float is narrowed before any module
+        // reaches a backend (ShaderTranspiler::DemoteFloat64Pass), so there is no `double` input
+        // left to bitcast INTO, and feeding a UINT-formatted attribute to what is now a `float`
+        // input would be silent garbage. Reconstructing the value would mean decoding the
+        // IEEE-754 double bit pattern in the shader - software fp64, which is precisely what the
+        // demotion exists to avoid - and on Espryt it would additionally need the ES driver to
+        // fetch 2N uint components where the application declared N doubles, which a dvec3 or
+        // dvec4 cannot even express within one attribute location.
+        //
+        // So glVertexAttribLFormat / glVertexAttribLPointer are declined here exactly as they
+        // already were on Espryt and on every real mobile device (Adreno and Mali both report
+        // shaderFloat64 == VK_FALSE), and for the same visible reason. A `dvec3` INPUT still
+        // compiles and draws - it is a `vec3` after demotion - as long as the application feeds
+        // it with glVertexAttribPointer(GL_FLOAT) rather than 64-bit data.
+        m_dynamicParameters.SupportsFloat64VertexAttributes = false;
         m_dynamicParameters.MaxShaderStorageBlockSize =
             std::min(m_vulkanCaps.MaxShaderStorageBlockSize, kMaxAdvertisedShaderStorageBlockSize);
         if (m_vulkanCaps.SupportsShaderSubgroup) {
@@ -818,6 +948,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             m_dynamicParameters.SubgroupSupportedFeatures =
                 mapSubgroupFeatures(m_vulkanCaps.SubgroupSupportedOperations);
             m_dynamicParameters.SubgroupQuadOperationsInAllStages = m_vulkanCaps.SubgroupQuadOperationsInAllStages;
+        } else if (ShouldEmulateSubgroups(m_vulkanCaps.SupportsShaderSubgroup)) {
+            // MOBILEGL_MAGMA_EMULATE_SUBGROUP on a device with no native subgroups: the
+            // advertised values describe the 32-lane virtual subgroup the compute
+            // lowering implements (SubgroupSupportPolicy.h / EmulateSubgroupsPass).
+            // GL requires the advertisement and the execution to agree, and on this
+            // path the emulation is what executes; only the compute stage is offered.
+            m_dynamicParameters.SubgroupSize = kEmulatedSubgroupSize;
+            m_dynamicParameters.SubgroupSupportedStages = kEmulatedSubgroupStages;
+            m_dynamicParameters.SubgroupSupportedFeatures = kEmulatedSubgroupFeatures;
+            m_dynamicParameters.SubgroupQuadOperationsInAllStages = false;
+            MGLOG_I("DirectVulkan: emulating 32-lane compute subgroups "
+                    "(MOBILEGL_MAGMA_EMULATE_SUBGROUP, no native subgroup support)");
         } else {
             m_dynamicParameters.SubgroupSize = 0;
             m_dynamicParameters.SubgroupSupportedStages = 0;

MobileGL/MG_Backend/DirectVulkan/BackendObject_DirectVulkan.h

@@ -62,8 +62,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     // POST screen shows.
 
     // Static identity of the Magma renderer (renderer/backend names, target GL/GLSL
-    // versions, ExtraVendor) with the baseline extension advertisement (no shader
+    // versions, ExtraVendor) with the baseline extension advertisement (no runtime-gated
-    // subgroup, no timer queries). A live backend copies this in its constructor and
+    // capabilities). A live backend copies this in its constructor and
     // reconciles the Extensions in UpdateAdvertisedExtensions once real capabilities
     // exist; callers that need the advertised list for a known capability set must
     // use BuildAdvertisedExtensions instead.
@@ -74,7 +74,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     // MOBILEGL_DISABLE_TIMERQUERY escape hatches are applied inside, so callers pass
     // the detected device support (passing an already-gated value is harmless).
     Vector<GLExtension> BuildAdvertisedExtensions(Bool shaderSubgroupSupported, Bool timerQueriesSupported,
-                                                  Bool anisotropicFilteringSupported);
+                                                  Bool anisotropicFilteringSupported,
+                                                  Bool nonZeroIndirectBaseInstanceSupported);
 
     // Format: <GPU Name>, Vulkan <Vulkan Version>, Driver <Driver Version> — the exact
     // string an initialized backend returns from GetBackendAPIVersionString (and that

MobileGL/MG_Backend/DirectVulkan/DirectVulkan.cpp

@@ -16,6 +16,7 @@
 #include "MG_Util/Metrics/TextureMetrics.h"
 #include "MG_Util/Miscellany/IndexGenerator.h"
 #include <atomic>
+#include <bit>
 #include <cstring>
 #include <spirv_reflect.h>
 
@@ -68,6 +69,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             // slot's ownership unambiguous.
             Uint64 programLifetimeId = 0;
             Uint32 backendStateVersion = 0;
+            // glShaderStorageBlockBinding deliberately does NOT bump the backend state
+            // version, and the pipeline composite is unnamed so the in-place patch in
+            // DirectVulkan::ShaderStorageBlockBinding can never reach its slot - the
+            // mirror replay bumps only the program's block-binding version. Without this
+            // key the composite's slot kept serving the pre-rebind block.binding.
+            Uint32 blockBindingVersion = 0;
             Vector<StorageBlockResource> storageBlocks;
             Vector<BufferVariableResource> bufferVariables;
             GLint computeWorkGroupSize[3] = {1, 1, 1};
@@ -155,18 +162,33 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             auto& cache = g_programResourceCaches[program.GetExternalIndex()];
             const Uint64 programLifetimeId = program.GetLifetimeId();
             const Uint32 backendStateVersion = program.GetBackendStateVersion();
+            const Uint32 blockBindingVersion = program.GetBlockBindingVersion();
             // The lifetime id must match too: a new program that reuses a deleted
             // program's name and happens to land on the same backendStateVersion (both
             // count from zero) would otherwise be served the dead program's reflection.
             if (cache.programLifetimeId == programLifetimeId &&
                 cache.backendStateVersion == backendStateVersion &&
                 (!cache.storageBlocks.empty() || !cache.bufferVariables.empty())) {
+                if (cache.blockBindingVersion != blockBindingVersion) {
+                    // Only the block bindings moved (glShaderStorageBlockBinding, or the
+                    // pipeline composite's mirror replay - neither touches the backend
+                    // state version): the reflection itself is unchanged, so re-apply the
+                    // overrides by name instead of re-running spirv-reflect. Overrides
+                    // only ever accumulate, so a block without one still holds its
+                    // declared binding.
+                    for (auto& block : cache.storageBlocks) {
+                        const Int rebound = program.GetShaderStorageBlockBindingOverride(block.name);
+                        if (rebound >= 0) block.binding = static_cast<Uint32>(rebound);
+                    }
+                    cache.blockBindingVersion = blockBindingVersion;
+                }
                 return cache;
             }
 
             cache = {};
             cache.programLifetimeId = programLifetimeId;
             cache.backendStateVersion = backendStateVersion;
+            cache.blockBindingVersion = blockBindingVersion;
 
             Vector<SpvReflectShaderModule> modules;
             Vector<Bool> validModules;
@@ -231,6 +253,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                     StorageBlockResource block{};
                     block.name = blockName;
                     block.binding = binding->binding;
+                    // glShaderStorageBlockBinding survives every rebuild of this cache: the
+                    // authoritative record of a rebound block lives on the program (it is what
+                    // GL_BUFFER_BINDING reports), and only the shader's declared binding is
+                    // recoverable from the SPIR-V. Without this, any unrelated state-version
+                    // bump would silently revert the block to its declared binding.
+                    const Int rebound = program.GetShaderStorageBlockBindingOverride(blockName);
+                    if (rebound >= 0) block.binding = static_cast<Uint32>(rebound);
                     block.dataSize = static_cast<GLint>(binding->block.size);
                     const GLuint blockIndex = static_cast<GLuint>(cache.storageBlocks.size());
                     AddBufferVariablesRecursive(binding->block, blockName, blockIndex, cache.bufferVariables,
@@ -255,132 +284,26 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return programObject.get();
         }
 
-        void CopyResourceName(const String& source, GLsizei bufSize, GLsizei* length, GLchar* name) {
-            const GLsizei writtenLength = static_cast<GLsizei>(source.size());
-            if (length) {
-                *length = writtenLength;
-            }
-            if (name && bufSize > 0) {
-                const GLsizei copyLength = std::min<GLsizei>(bufSize - 1, writtenLength);
-                std::memcpy(name, source.data(), static_cast<SizeT>(copyLength));
-                name[copyLength] = '\0';
-            }
-        }
-
         const Uint8* ResolveIndirectCommandBytes(const void* indirect, SizeT requiredBytes, const char* label) {
             auto drawBuffer = MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DrawIndirect).GetBoundObject();
             if (drawBuffer) {
                 drawBuffer->SyncPersistentMappedRange();
                 const SizeT commandOffset = reinterpret_cast<SizeT>(indirect);
                 if (drawBuffer->MappedData() == nullptr || commandOffset + requiredBytes > drawBuffer->GetSize()) {
-                    MGLOG_E("%s skipped: invalid GL_DRAW_INDIRECT_BUFFER binding or range", label);
+                    MGLOG_E_ONCE("%s skipped: invalid GL_DRAW_INDIRECT_BUFFER binding or range", label);
                     return nullptr;
                 }
                 return drawBuffer->MappedData() + commandOffset;
             }
 
             if (!indirect) {
-                MGLOG_E("%s skipped: indirect pointer is null", label);
+                MGLOG_E_ONCE("%s skipped: indirect pointer is null", label);
                 return nullptr;
             }
 
             return reinterpret_cast<const Uint8*>(indirect);
         }
 
-        Vector<GLuint> GetUniformBlockActiveVariables(const MG_State::GLState::ProgramObject& program,
-                                                      GLuint blockIndex) {
-            Vector<GLuint> activeVariables;
-            const Uint uniformCount = program.GetUniformCount();
-            activeVariables.reserve(uniformCount);
-            for (Uint uniformIndex = 0; uniformIndex < uniformCount; ++uniformIndex) {
-                if (program.GetActiveUniformBlockIndex(uniformIndex) == static_cast<Int>(blockIndex)) {
-                    activeVariables.push_back(uniformIndex);
-                }
-            }
-            return activeVariables;
-        }
-
-        GLuint FindProgramInputIndex(const MG_State::GLState::ProgramObject& program, const String& name) {
-            const Int activeCount = program.GetActiveAttributesCount();
-            for (Int index = 0; index < activeCount; ++index) {
-                if (program.GetActiveAttribName(index) == name) {
-                    return static_cast<GLuint>(index);
-                }
-            }
-            return GL_INVALID_INDEX;
-        }
-
-        GLuint FindProgramOutputIndex(const MG_State::GLState::ProgramObject& program, const String& name) {
-            const Int activeCount = program.GetActiveFragmentOutputCount();
-            for (Int index = 0; index < activeCount; ++index) {
-                if (program.GetActiveFragmentOutputName(index) == name) {
-                    return static_cast<GLuint>(index);
-                }
-            }
-            return GL_INVALID_INDEX;
-        }
-
-        GLint GetProgramOutputLocation(const MG_State::GLState::ProgramObject& program, const String& name) {
-            const Int activeCount = program.GetActiveFragmentOutputCount();
-            for (Int index = 0; index < activeCount; ++index) {
-                if (program.GetActiveFragmentOutputName(index) == name) {
-                    return program.GetFragmentOutputLocation(index);
-                }
-            }
-            return -1;
-        }
-
-        GLint GetProgramResourceActiveCount(const MG_State::GLState::ProgramObject& program, GLenum programInterface,
-                                            const ProgramResourceCache& cache) {
-            switch (programInterface) {
-            case GL_SHADER_STORAGE_BLOCK:
-                return static_cast<GLint>(cache.storageBlocks.size());
-            case GL_BUFFER_VARIABLE:
-                return static_cast<GLint>(cache.bufferVariables.size());
-            case GL_UNIFORM_BLOCK:
-                return program.GetActiveUniformBlocksCount();
-            case GL_UNIFORM:
-                return static_cast<GLint>(program.GetUniformCount());
-            case GL_PROGRAM_INPUT:
-                return program.GetActiveAttributesCount();
-            case GL_PROGRAM_OUTPUT:
-                return program.GetActiveFragmentOutputCount();
-            default:
-                return 0;
-            }
-        }
-
-        GLint GetProgramResourceMaxNameLength(const MG_State::GLState::ProgramObject& program, GLenum programInterface,
-                                              const ProgramResourceCache& cache) {
-            switch (programInterface) {
-            case GL_SHADER_STORAGE_BLOCK: {
-                SizeT maxLength = 0;
-                for (const auto& block : cache.storageBlocks) maxLength = std::max(maxLength, block.name.size() + 1);
-                return static_cast<GLint>(maxLength);
-            }
-            case GL_BUFFER_VARIABLE: {
-                SizeT maxLength = 0;
-                for (const auto& var : cache.bufferVariables) maxLength = std::max(maxLength, var.name.size() + 1);
-                return static_cast<GLint>(maxLength);
-            }
-            case GL_UNIFORM_BLOCK:
-                return program.GetActiveUniformBlocksMaxNameLength() + 1;
-            case GL_UNIFORM:
-                return program.GetUniformMaxLength() + 1;
-            case GL_PROGRAM_INPUT:
-                return program.GetActiveAttributesMaxLength() + 1;
-            case GL_PROGRAM_OUTPUT: {
-                SizeT maxLength = 0;
-                const Int activeCount = program.GetActiveFragmentOutputCount();
-                for (Int index = 0; index < activeCount; ++index) {
-                    maxLength = std::max(maxLength, program.GetActiveFragmentOutputName(index).size() + 1);
-                }
-                return static_cast<GLint>(maxLength);
-            }
-            default:
-                return 0;
-            }
-        }
     } // namespace
 
     void ClearProgramResourceCaches() {
@@ -394,11 +317,21 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
     GLuint GetShaderStorageBlockIndex(const MG_State::GLState::ProgramObject& program, const String& name) {
         auto& cache = GetProgramResourceCache(program);
-        const auto it = std::find_if(cache.storageBlocks.begin(), cache.storageBlocks.end(),
+        auto find = [&cache](const String& key) {
-            [&](const StorageBlockResource& block) { return block.name == name; });
+            return std::find_if(cache.storageBlocks.begin(), cache.storageBlocks.end(),
-        return it == cache.storageBlocks.end()
+                [&](const StorageBlockResource& block) { return block.name == key; });
-            ? GL_INVALID_INDEX
+        };
-            : static_cast<GLuint>(std::distance(cache.storageBlocks.begin(), it));
+        auto it = find(name);
+        if (it == cache.storageBlocks.end()) {
+            // Cache names are normalized (NormalizeDescriptorName drops the array suffix), so
+            // an arrayed block that GL enumerates per element - "B[0]", "B[1]" - is one entry
+            // here, spelled "B". Retry against the bare name before giving up.
+            const auto bracket = name.rfind('[');
+            if (bracket == String::npos || name.empty() || name.back() != ']') return GL_INVALID_INDEX;
+            it = find(name.substr(0, bracket));
+            if (it == cache.storageBlocks.end()) return GL_INVALID_INDEX;
+        }
+        return static_cast<GLuint>(std::distance(cache.storageBlocks.begin(), it));
     }
 
     GLuint GetShaderStorageBlockBinding(const MG_State::GLState::ProgramObject& program, GLuint blockIndex) {
@@ -486,7 +419,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             stride = sizeof(DrawArraysIndirectCommand);
         }
         if (stride < static_cast<GLsizei>(sizeof(DrawArraysIndirectCommand))) {
-            MGLOG_E("MultiDrawArraysIndirect skipped: stride %d is smaller than command size %zu",
+            MGLOG_E_ONCE("MultiDrawArraysIndirect skipped: stride %d is smaller than command size %zu",
                     stride, sizeof(DrawArraysIndirectCommand));
             return;
         }
@@ -534,20 +467,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             stride = sizeof(DrawArraysIndirectCommand);
         }
         if (stride < static_cast<GLsizei>(sizeof(DrawArraysIndirectCommand))) {
-            MGLOG_E("MultiDrawArraysIndirectCount skipped: stride %d is smaller than command size %zu",
+            MGLOG_E_ONCE("MultiDrawArraysIndirectCount skipped: stride %d is smaller than command size %zu",
                     stride, sizeof(DrawArraysIndirectCommand));
             return;
         }
 
         auto parameterBuffer = MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::Parameter).GetBoundObject();
         if (!parameterBuffer || drawcount < 0 || static_cast<SizeT>(drawcount) + sizeof(Uint32) > parameterBuffer->GetSize()) {
-            MGLOG_E("MultiDrawArraysIndirectCount skipped: invalid GL_PARAMETER_BUFFER binding or range");
+            MGLOG_E_ONCE("MultiDrawArraysIndirectCount skipped: invalid GL_PARAMETER_BUFFER binding or range");
             return;
         }
 
         parameterBuffer->SyncPersistentMappedRange();
         if (parameterBuffer->MappedData() == nullptr) {
-            MGLOG_E("MultiDrawArraysIndirectCount skipped: CPU fallback cannot read parameter buffer");
+            MGLOG_E_ONCE("MultiDrawArraysIndirectCount skipped: CPU fallback cannot read parameter buffer");
             return;
         }
 
@@ -601,7 +534,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         const SizeT indexSize = MG_Util::GetGLTypeSize(type);
         if (indexSize == 0) {
-            MGLOG_E("DrawElementsIndirect skipped: unsupported index type 0x%x", type);
+            MGLOG_E_ONCE("DrawElementsIndirect skipped: unsupported index type 0x%x", type);
             return;
         }
 
@@ -873,357 +806,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
     }
 
-    void GetProgramInterfaceiv(GLuint program, GLenum programInterface, GLenum pname, GLint* params) {
+    void ShaderStorageBlockBinding(GLuint program, const GLchar* storageBlockName, GLuint storageBlockBinding) {
-        if (!params) return;
         auto* programObject = TryGetDirectVulkanProgram(program);
-        if (!programObject) return;
+        if (!programObject || storageBlockName == nullptr) return;
-        auto& cache = GetProgramResourceCache(*programObject);
-        switch (pname) {
-        case GL_ACTIVE_RESOURCES:
-            *params = GetProgramResourceActiveCount(*programObject, programInterface, cache);
-            return;
-        case GL_MAX_NAME_LENGTH:
-            *params = GetProgramResourceMaxNameLength(*programObject, programInterface, cache);
-            return;
-        case GL_MAX_NUM_ACTIVE_VARIABLES:
-            if (programInterface == GL_SHADER_STORAGE_BLOCK) {
-                SizeT maxCount = 0;
-                for (const auto& block : cache.storageBlocks) {
-                    maxCount = std::max(maxCount, block.activeVariables.size());
-                }
-                *params = static_cast<GLint>(maxCount);
-            } else if (programInterface == GL_UNIFORM_BLOCK) {
-                GLint maxCount = 0;
-                const Int activeBlocks = programObject->GetActiveUniformBlocksCount();
-                for (Int index = 0; index < activeBlocks; ++index) {
-                    maxCount = std::max(maxCount, programObject->GetUniformBlockActiveUniformCount(index));
-                }
-                *params = maxCount;
-            } else {
-                *params = 0;
-            }
-            return;
-        default:
-            *params = 0;
-            return;
-        }
-    }
-
-    GLuint GetProgramResourceIndex(GLuint program, GLenum programInterface, const GLchar* name) {
-        if (!name) return GL_INVALID_INDEX;
-        auto* programObject = TryGetDirectVulkanProgram(program);
-        if (!programObject) return GL_INVALID_INDEX;
-        auto& cache = GetProgramResourceCache(*programObject);
-        const String resourceName = name;
-        if (programInterface == GL_SHADER_STORAGE_BLOCK) {
-            return GetShaderStorageBlockIndex(*programObject, name);
-        }
-        if (programInterface == GL_BUFFER_VARIABLE) {
-            const auto it = std::find_if(cache.bufferVariables.begin(), cache.bufferVariables.end(),
-                [&](const BufferVariableResource& var) { return var.name == resourceName; });
-            return it == cache.bufferVariables.end()
-                ? GL_INVALID_INDEX
-                : static_cast<GLuint>(std::distance(cache.bufferVariables.begin(), it));
-        }
-        if (programInterface == GL_UNIFORM_BLOCK) {
-            return programObject->GetUniformBlockIndex(name);
-        }
-        if (programInterface == GL_UNIFORM) {
-            const Int activeUniformIndex = programObject->GetActiveUniformIndex(resourceName);
-            return activeUniformIndex >= 0 ? static_cast<GLuint>(activeUniformIndex) : GL_INVALID_INDEX;
-        }
-        if (programInterface == GL_PROGRAM_INPUT) {
-            return FindProgramInputIndex(*programObject, resourceName);
-        }
-        if (programInterface == GL_PROGRAM_OUTPUT) {
-            return FindProgramOutputIndex(*programObject, resourceName);
-        }
-        return GL_INVALID_INDEX;
-    }
-
-    void GetProgramResourceName(GLuint program, GLenum programInterface, GLuint index, GLsizei bufSize,
-                                GLsizei* length, GLchar* name) {
-        auto* programObject = TryGetDirectVulkanProgram(program);
-        if (!programObject) return;
-        auto& cache = GetProgramResourceCache(*programObject);
-        if (programInterface == GL_SHADER_STORAGE_BLOCK && index < cache.storageBlocks.size()) {
-            CopyResourceName(cache.storageBlocks[index].name, bufSize, length, name);
-            return;
-        }
-        if (programInterface == GL_BUFFER_VARIABLE && index < cache.bufferVariables.size()) {
-            CopyResourceName(cache.bufferVariables[index].name, bufSize, length, name);
-            return;
-        }
-        if (programInterface == GL_UNIFORM_BLOCK && programObject->IsActiveUniformBlock(index)) {
-            CopyResourceName(programObject->GetUniformBlockName(index), bufSize, length, name);
-            return;
-        }
-        if (programInterface == GL_UNIFORM && index < programObject->GetUniformCount()) {
-            CopyResourceName(programObject->GetActiveUniformName(index), bufSize, length, name);
-            return;
-        }
-        if (programInterface == GL_PROGRAM_INPUT && index < static_cast<GLuint>(programObject->GetActiveAttributesCount())) {
-            CopyResourceName(programObject->GetActiveAttribName(index), bufSize, length, name);
-            return;
-        }
-        if (programInterface == GL_PROGRAM_OUTPUT &&
-            index < static_cast<GLuint>(programObject->GetActiveFragmentOutputCount())) {
-            CopyResourceName(programObject->GetActiveFragmentOutputName(index), bufSize, length, name);
-            return;
-        }
-        if (length) *length = 0;
-        if (name && bufSize > 0) name[0] = '\0';
-    }
-
-    void GetProgramResourceiv(GLuint program, GLenum programInterface, GLuint index, GLsizei propCount,
-                              const GLenum* props, GLsizei bufSize, GLsizei* length, GLint* params) {
-        auto* programObject = TryGetDirectVulkanProgram(program);
-        if (!programObject || !props || !params || bufSize <= 0) return;
-        auto& cache = GetProgramResourceCache(*programObject);
-        GLsizei written = 0;
-        auto writeValue = [&](GLint value) {
-            if (written < bufSize) {
-                params[written++] = value;
-            }
-        };
-
-        for (GLsizei propIndex = 0; propIndex < propCount; ++propIndex) {
-            const GLenum prop = props[propIndex];
-            if (programInterface == GL_SHADER_STORAGE_BLOCK && index < cache.storageBlocks.size()) {
-                const auto& block = cache.storageBlocks[index];
-                switch (prop) {
-                case GL_NAME_LENGTH:
-                    writeValue(static_cast<GLint>(block.name.size() + 1));
-                    break;
-                case GL_BUFFER_BINDING:
-                    writeValue(static_cast<GLint>(block.binding));
-                    break;
-                case GL_BUFFER_DATA_SIZE:
-                    writeValue(block.dataSize);
-                    break;
-                case GL_NUM_ACTIVE_VARIABLES:
-                    writeValue(static_cast<GLint>(block.activeVariables.size()));
-                    break;
-                case GL_ACTIVE_VARIABLES:
-                    for (const auto variable : block.activeVariables) writeValue(static_cast<GLint>(variable));
-                    break;
-                default:
-                    writeValue(0);
-                    break;
-                }
-            } else if (programInterface == GL_BUFFER_VARIABLE && index < cache.bufferVariables.size()) {
-                const auto& var = cache.bufferVariables[index];
-                switch (prop) {
-                case GL_NAME_LENGTH:
-                    writeValue(static_cast<GLint>(var.name.size() + 1));
-                    break;
-                case GL_TYPE:
-                    writeValue(GL_FLOAT);
-                    break;
-                case GL_ARRAY_SIZE:
-                    writeValue(1);
-                    break;
-                case GL_OFFSET:
-                    writeValue(var.offset);
-                    break;
-                case GL_BLOCK_INDEX:
-                    writeValue(static_cast<GLint>(var.blockIndex));
-                    break;
-                case GL_ARRAY_STRIDE:
-                case GL_MATRIX_STRIDE:
-                case GL_TOP_LEVEL_ARRAY_SIZE:
-                case GL_TOP_LEVEL_ARRAY_STRIDE:
-                case GL_IS_ROW_MAJOR:
-                    writeValue(0);
-                    break;
-                default:
-                    writeValue(0);
-                    break;
-                }
-            } else if (programInterface == GL_UNIFORM_BLOCK &&
-                       programObject->IsActiveUniformBlock(index)) {
-                const auto activeVariables = GetUniformBlockActiveVariables(*programObject, index);
-                switch (prop) {
-                case GL_NAME_LENGTH:
-                    writeValue(static_cast<GLint>(programObject->GetUniformBlockName(index).size() + 1));
-                    break;
-                case GL_BUFFER_BINDING:
-                    writeValue(static_cast<GLint>(programObject->GetUniformBlockBinding(index)));
-                    break;
-                case GL_BUFFER_DATA_SIZE:
-                    writeValue(static_cast<GLint>(programObject->GetUBOSizeAt(index)));
-                    break;
-                case GL_NUM_ACTIVE_VARIABLES:
-                    writeValue(static_cast<GLint>(activeVariables.size()));
-                    break;
-                case GL_ACTIVE_VARIABLES:
-                    for (const GLuint variableIndex : activeVariables) {
-                        writeValue(static_cast<GLint>(variableIndex));
-                    }
-                    break;
-                case GL_REFERENCED_BY_VERTEX_SHADER:
-                    writeValue(programObject->IsUniformBlockReferencedByStage(index, EShLangVertex) ? GL_TRUE
-                                                                                                     : GL_FALSE);
-                    break;
-                case GL_REFERENCED_BY_FRAGMENT_SHADER:
-                    writeValue(programObject->IsUniformBlockReferencedByStage(index, EShLangFragment) ? GL_TRUE
-                                                                                                       : GL_FALSE);
-                    break;
-                case GL_REFERENCED_BY_COMPUTE_SHADER:
-                    writeValue(programObject->IsUniformBlockReferencedByStage(index, EShLangCompute) ? GL_TRUE
-                                                                                                      : GL_FALSE);
-                    break;
-                case GL_REFERENCED_BY_GEOMETRY_SHADER:
-                case GL_REFERENCED_BY_TESS_CONTROL_SHADER:
-                case GL_REFERENCED_BY_TESS_EVALUATION_SHADER:
-                    writeValue(GL_FALSE);
-                    break;
-                default:
-                    writeValue(0);
-                    break;
-                }
-            } else if (programInterface == GL_UNIFORM && index < programObject->GetUniformCount()) {
-                const auto& uniformName = programObject->GetActiveUniformName(index);
-                const GLint location = programObject->GetUniformLocation(uniformName);
-                switch (prop) {
-                case GL_NAME_LENGTH:
-                    writeValue(static_cast<GLint>(uniformName.size() + 1));
-                    break;
-                case GL_TYPE:
-                    writeValue(static_cast<GLint>(programObject->GetActiveUniformType(index)));
-                    break;
-                case GL_ARRAY_SIZE:
-                    writeValue(programObject->GetActiveUniformArraySize(index));
-                    break;
-                case GL_BLOCK_INDEX:
-                    writeValue(programObject->GetActiveUniformBlockIndex(index));
-                    break;
-                case GL_LOCATION:
-                    writeValue(location);
-                    break;
-                case GL_OFFSET:
-                    writeValue(location >= 0 && programObject->IsValidUniformLocation(location)
-                                   ? static_cast<GLint>(programObject->GetUniformOffset(location))
-                                   : 0);
-                    break;
-                case GL_ARRAY_STRIDE:
-                case GL_MATRIX_STRIDE:
-                case GL_IS_ROW_MAJOR:
-                case GL_TOP_LEVEL_ARRAY_SIZE:
-                case GL_TOP_LEVEL_ARRAY_STRIDE:
-                case GL_REFERENCED_BY_VERTEX_SHADER:
-                case GL_REFERENCED_BY_FRAGMENT_SHADER:
-                case GL_REFERENCED_BY_COMPUTE_SHADER:
-                case GL_REFERENCED_BY_GEOMETRY_SHADER:
-                case GL_REFERENCED_BY_TESS_CONTROL_SHADER:
-                case GL_REFERENCED_BY_TESS_EVALUATION_SHADER:
-                    writeValue(0);
-                    break;
-                default:
-                    writeValue(0);
-                    break;
-                }
-            } else if (programInterface == GL_PROGRAM_INPUT &&
-                       index < static_cast<GLuint>(programObject->GetActiveAttributesCount())) {
-                const auto& resourceName = programObject->GetActiveAttribName(index);
-                switch (prop) {
-                case GL_NAME_LENGTH:
-                    writeValue(static_cast<GLint>(resourceName.size() + 1));
-                    break;
-                case GL_TYPE:
-                    writeValue(static_cast<GLint>(programObject->GetActiveAttribType(index)));
-                    break;
-                case GL_ARRAY_SIZE:
-                    writeValue(programObject->GetActiveAttribArraySize(index));
-                    break;
-                case GL_LOCATION:
-                    writeValue(programObject->GetAttributeLocation(resourceName));
-                    break;
-                case GL_REFERENCED_BY_VERTEX_SHADER:
-                    writeValue(GL_TRUE);
-                    break;
-                case GL_REFERENCED_BY_FRAGMENT_SHADER:
-                case GL_REFERENCED_BY_COMPUTE_SHADER:
-                case GL_REFERENCED_BY_GEOMETRY_SHADER:
-                case GL_REFERENCED_BY_TESS_CONTROL_SHADER:
-                case GL_REFERENCED_BY_TESS_EVALUATION_SHADER:
-                case GL_IS_PER_PATCH:
-                case GL_LOCATION_INDEX:
-                    writeValue(0);
-                    break;
-                default:
-                    writeValue(0);
-                    break;
-                }
-            } else if (programInterface == GL_PROGRAM_OUTPUT &&
-                       index < static_cast<GLuint>(programObject->GetActiveFragmentOutputCount())) {
-                const auto& resourceName = programObject->GetActiveFragmentOutputName(index);
-                switch (prop) {
-                case GL_NAME_LENGTH:
-                    writeValue(static_cast<GLint>(resourceName.size() + 1));
-                    break;
-                case GL_TYPE:
-                    writeValue(static_cast<GLint>(programObject->GetFragmentOutputType(index)));
-                    break;
-                case GL_ARRAY_SIZE:
-                    writeValue(programObject->GetActiveFragmentOutputArraySize(index));
-                    break;
-                case GL_LOCATION:
-                    writeValue(programObject->GetFragmentOutputLocation(index));
-                    break;
-                case GL_LOCATION_INDEX:
-                    writeValue(0);
-                    break;
-                case GL_REFERENCED_BY_FRAGMENT_SHADER:
-                    writeValue(GL_TRUE);
-                    break;
-                case GL_REFERENCED_BY_VERTEX_SHADER:
-                case GL_REFERENCED_BY_COMPUTE_SHADER:
-                case GL_REFERENCED_BY_GEOMETRY_SHADER:
-                case GL_REFERENCED_BY_TESS_CONTROL_SHADER:
-                case GL_REFERENCED_BY_TESS_EVALUATION_SHADER:
-                case GL_IS_PER_PATCH:
-                    writeValue(0);
-                    break;
-                default:
-                    writeValue(0);
-                    break;
-                }
-            } else {
-                writeValue(0);
-            }
-        }
-        if (length) *length = written;
-    }
-
-    GLint GetProgramResourceLocation(GLuint program, GLenum programInterface, const GLchar* name) {
-        auto* programObject = TryGetDirectVulkanProgram(program);
-        if (!programObject || !name) return -1;
-        if (programInterface == GL_UNIFORM) {
-            return programObject->GetUniformLocation(name);
-        }
-        if (programInterface == GL_PROGRAM_INPUT) {
-            return programObject->GetAttributeLocation(name);
-        }
-        if (programInterface == GL_PROGRAM_OUTPUT) {
-            return GetProgramOutputLocation(*programObject, name);
-        }
-        return -1;
-    }
-
-    GLint GetProgramResourceLocationIndex(GLuint program, GLenum programInterface, const GLchar* name) {
-        auto* programObject = TryGetDirectVulkanProgram(program);
-        if (!programObject || !name) return -1;
-        if (programInterface == GL_PROGRAM_OUTPUT) {
-            return GetProgramOutputLocation(*programObject, name) >= 0 ? 0 : -1;
-        }
-        return -1;
-    }
-
-    void ShaderStorageBlockBinding(GLuint program, GLuint storageBlockIndex, GLuint storageBlockBinding) {
-        auto* programObject = TryGetDirectVulkanProgram(program);
-        if (!programObject) return;
-        auto& cache = GetProgramResourceCache(*programObject);
         const Int maxBindings = pActiveBackendObject
             ? pActiveBackendObject->GetDynamicParameters().MaxShaderStorageBufferBindings
             : 0;
@@ -1233,13 +818,23 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 MakeUnique<GenericErrorInfo>("DirectVulkan", __func__, "Shader storage binding is out of range."));
             return;
         }
-        if (storageBlockIndex >= cache.storageBlocks.size()) {
+        // The frontend already validated that the name denotes an active block, and has
-            MG_State::pGLContext->RecordError(
+        // already recorded the new binding on the program - which is what reseeds this cache
-                ErrorCode::InvalidValue,
+        // whenever it is rebuilt. Writing the entry here as well keeps an ALREADY-BUILT cache
-                MakeUnique<GenericErrorInfo>("DirectVulkan", __func__, "Shader storage block index is not active."));
+        // (the common case: the very next draw reads it) from having to be thrown away.
-            return;
+        //
-        }
+        // Resolve the index BEFORE taking the reference, and bounds-check the way the
-        cache.storageBlocks[storageBlockIndex].binding = storageBlockBinding;
+        // sibling getter does. GetShaderStorageBlockIndex re-enters GetProgramResourceCache,
+        // which indexes g_programResourceCaches and can therefore insert - and that map is
+        // open-addressed, so a rehash MOVES its entries and a reference taken before the
+        // call is left dangling. Binding a program's storage block
+        // while another program's entry was still absent from the cache was a reproducible
+        // segfault (ProgramPipelineScenario's two storage-block cases, in one process).
+        const GLuint blockIndex = GetShaderStorageBlockIndex(*programObject, storageBlockName);
+        if (blockIndex == GL_INVALID_INDEX) return;
+        auto& cache = GetProgramResourceCache(*programObject);
+        if (blockIndex >= cache.storageBlocks.size()) return;
+        cache.storageBlocks[blockIndex].binding = storageBlockBinding;
     }
     void ReadPixels(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, void* pixels) {
         MOBILEGL_ASSERT(pVulkanRenderer, "DirectVulkan::ReadPixels called with null VulkanRenderer");
@@ -1392,31 +987,89 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         pVulkanRenderer->MultiDrawArrays(payload);
     }
 
+    // Shared body of glMultiDrawElements (basevertex == nullptr) and
+    // glMultiDrawElementsBaseVertex: identical calls except for the per-draw
+    // vertex offset, which VkMultiDrawIndexedInfoEXT / VkDrawIndexedIndirectCommand /
+    // vkCmdDrawIndexed all carry natively.
+    static void MultiDrawElementsImpl(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
+                                      GLsizei drawcount, const GLint* basevertex) {
+        if (drawcount <= 0) {
+            return;
+        }
+
+        // With no element-array buffer bound, every indices[i] is a client pointer into a
+        // separate CPU allocation, not an offset into one shared buffer. The batched payload
+        // below cannot express that: it carries ONE index-buffer view for the whole batch and
+        // turns each pointer into a firstIndex relative to it. Replay the sub-draws through
+        // the single-draw entry point instead - it snapshots each client range into its own
+        // transient slice, which is exactly what the unrolled draws this must match do.
+        // (The batch used to be built this way; the shared-view rewrite that added
+        // MultiDrawIndexedCmd left the client-memory shape addressing a view whose byte
+        // offset is a hardcoded 0, so UploadAndBindIndexBuffer saw a null client pointer,
+        // declined the whole batch and painted nothing.)
+        const auto& vao = *MG_State::pGLContext->GetBoundVertexArray();
+        if (vao.GetIndexBufferBindingSlot().GetBoundObject() == nullptr) {
+            for (GLsizei i = 0; i < drawcount; ++i) {
+                if (count[i] <= 0) {
+                    continue;
+                }
+                DrawElementsBaseVertex(mode, count[i], type, indices[i],
+                                       basevertex != nullptr ? basevertex[i] : 0);
+            }
+            return;
+        }
+
+        MultiDrawIndexedCmd payload{};
+        payload.mode = mode;
+        payload.indexBufferView.indexType = type;
+
+        // Loop-invariant: the index type is fixed for the whole multi-draw, so resolve
+        // its byte size once instead of twice per sub-draw (a cross-TU switch that
+        // showed up in per-frame profiles of sodium-style 132x32 multi-draws). Index
+        // sizes are 1/2/4, so the per-sub-draw offset division below reduces to a
+        // shift - the hardware divide was the hottest instruction of this loop.
+        const SizeT indexSize = MG_Util::GetGLTypeSize(type);
+        if (indexSize == 0) {
+            MGLOG_E_ONCE("MultiDrawElements skipped: unsupported index type 0x%x", type);
+            return;
+        }
+        const Uint32 indexSizeShift = static_cast<Uint32>(std::countr_zero(indexSize));
+
+        // TODO: allocate draw cmd buf elsewhere
+        static Vector<DrawIndexedCmdParam> params;
+        params.clear();
+        params.resize(drawcount);
+
+        for (GLsizei i = 0; i < drawcount; ++i) {
+            if (count[i] == 0) {
+                continue;
+            }
+
+            // TODO: this index view needs a redesign, now there's a lotta redundant uploads
+
+            payload.indexBufferView.indexByteOffset = 0;
+            payload.indexBufferView.indexByteSize =
+                    std::max(reinterpret_cast<SizeT>(indices[i]) + count[i] * indexSize,
+                             payload.indexBufferView.indexByteSize);
+
+            auto& param = params[i];
+
+            param.indexCount = count[i];
+            param.instanceCount = 1;
+            param.firstIndex = reinterpret_cast<SizeT>(indices[i]) >> indexSizeShift;
+            param.vertexOffset = basevertex != nullptr ? basevertex[i] : 0;
+            param.firstInstance = 0;
+        }
+        payload.drawCount = drawcount;
+        payload.pParams = params.data();
+        pVulkanRenderer->MultiDrawElements(payload);
+    }
+
     void MultiDrawElements(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
                            GLsizei drawcount) {
         MOBILEGL_ASSERT(pVulkanRenderer, "DirectVulkan::MultiDrawElements called with null VulkanRenderer");
         MOBILEGL_ASSERT(MG_State::pGLContext, "DirectVulkan::MultiDrawElements called with null GL context");
-
+        MultiDrawElementsImpl(mode, count, type, indices, drawcount, nullptr);
-        // Vector<DrawElementCmd> cmds;
-        // cmds.reserve(static_cast<SizeT>(drawcount));
-        // for (GLsizei i = 0; i < drawcount; ++i) {
-        //     if (count[i] == 0) {
-        //         continue;
-        //     }
-        //
-        //     DrawElementCmd payload{};
-        //     payload.mode = mode;
-        //     payload.firstVertex = 0;
-        //     payload.indexCount = count[i];
-        //     payload.indexType = type;
-        //     payload.indexByteOffset = reinterpret_cast<SizeT>(indices[i]);
-        //     cmds.push_back(payload);
-        // }
-        //
-        // if (cmds.empty()) {
-        //     return;
-        // }
-        // pVulkanRenderer->MultiDrawElements(cmds);
     }
 
     void DrawElementsBaseVertex(GLenum mode, GLsizei count, GLenum type, const GLvoid* indices, GLint basevertex) {
@@ -1444,40 +1097,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
     void MultiDrawElementsBaseVertex(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
                                      GLsizei drawcount, const GLint* basevertex) {
-        MOBILEGL_ASSERT(pVulkanRenderer, "DirectVulkan::MultiDrawElements called with null VulkanRenderer");
+        MOBILEGL_ASSERT(pVulkanRenderer, "DirectVulkan::MultiDrawElementsBaseVertex called with null VulkanRenderer");
-        MOBILEGL_ASSERT(MG_State::pGLContext, "DirectVulkan::MultiDrawElements called with null GL context");
+        MOBILEGL_ASSERT(MG_State::pGLContext, "DirectVulkan::MultiDrawElementsBaseVertex called with null GL context");
-        MultiDrawIndexedCmd payload{};
+        MultiDrawElementsImpl(mode, count, type, indices, drawcount, basevertex);
-        payload.mode = mode;
-        payload.indexBufferView.indexType = type;
-
-        // TODO: allocate draw cmd buf elsewhere
-        static Vector<DrawIndexedCmdParam> params;
-        params.clear();
-        params.resize(drawcount);
-
-        for (GLsizei i = 0; i < drawcount; ++i) {
-            if (count[i] == 0) {
-                continue;
-            }
-
-            // TODO: this index view needs a redesign, now there's a lotta redundant uploads
-
-            payload.indexBufferView.indexByteOffset = 0;
-            payload.indexBufferView.indexByteSize =
-                    std::max(reinterpret_cast<SizeT>(indices[i]) + count[i] * MG_Util::GetGLTypeSize(type),
-                             payload.indexBufferView.indexByteSize);
-
-            auto& param = params[i];
-
-            param.indexCount = count[i];
-            param.instanceCount = 1;
-            param.firstIndex = reinterpret_cast<SizeT>(indices[i]) / MG_Util::GetGLTypeSize(type);
-            param.vertexOffset = basevertex[i];
-            param.firstInstance = 0;
-        }
-        payload.drawCount = drawcount;
-        payload.pParams = params.data();
-        pVulkanRenderer->MultiDrawElements(payload);
     }
 
     void BlitFramebuffer(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1,

MobileGL/MG_Backend/DirectVulkan/DirectVulkan.h

@@ -97,15 +97,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     void GetIntegeri_v(GLenum target, GLuint index, GLint* data);
     void GetInteger64i_v(GLenum target, GLuint index, GLint64* data);
     void GetProgramiv(GLuint program, GLenum pname, GLint* params);
-    void GetProgramInterfaceiv(GLuint program, GLenum programInterface, GLenum pname, GLint* params);
+    void ShaderStorageBlockBinding(GLuint program, const GLchar* storageBlockName, GLuint storageBlockBinding);
-    GLuint GetProgramResourceIndex(GLuint program, GLenum programInterface, const GLchar* name);
-    void GetProgramResourceName(GLuint program, GLenum programInterface, GLuint index, GLsizei bufSize,
-                                GLsizei* length, GLchar* name);
-    void GetProgramResourceiv(GLuint program, GLenum programInterface, GLuint index, GLsizei propCount,
-                              const GLenum* props, GLsizei bufSize, GLsizei* length, GLint* params);
-    GLint GetProgramResourceLocation(GLuint program, GLenum programInterface, const GLchar* name);
-    GLint GetProgramResourceLocationIndex(GLuint program, GLenum programInterface, const GLchar* name);
-    void ShaderStorageBlockBinding(GLuint program, GLuint storageBlockIndex, GLuint storageBlockBinding);
     void ReadPixels(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, void* pixels);
     void GetTexImage(GLenum target, GLint level, GLenum format, GLenum type, GLvoid* pixels);
     void GetTextureImage(const SharedPtr<MG_State::GLState::ITextureObject>& texture, TextureUploadTarget uploadTarget,

MobileGL/MG_Backend/DirectVulkan/Renderer/BufferArena.cpp

@@ -111,6 +111,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
 
         if (buffer.IsValid()) {
+            // Outgrown, not dead: every BufferSlice handed out from this frame's arena so far
+            // still names it, and those slices stay in service until the frame slot is rewound
+            // (VkBufferResource::transientSlice, the converted-vertex-stream cache, the draw
+            // memos). The release therefore has to survive every mid-frame reclaim and land on
+            // the next ResetFrame of this slot - see VkBufferManager::CollectAllDeferredReleases.
             m_deferredReleases[frameIndex].push_back(std::move(buffer));
         }
 

MobileGL/MG_Backend/DirectVulkan/Renderer/FrameContext.cpp

@@ -205,7 +205,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // commands away. The device is gone on that path anyway - stay silent-safe
         // rather than trade a lost device for a barrier into a closed buffer.
         if (frame.hasCommandBufferRecorded) {
-            MGLOG_E("TransitionToPresent: command buffer already closed; skipping the present barrier");
+            MGLOG_E_ONCE("TransitionToPresent: command buffer already closed; skipping the present barrier");
             return false;
         }
 

MobileGL/MG_Backend/DirectVulkan/Renderer/PipelineFactory.cpp

@@ -206,9 +206,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         XXHASH_VERIFY(
             XXH64_update(m_hashState, &payload.primitiveRestartEnable, sizeof(payload.primitiveRestartEnable)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.patchControlPoints, sizeof(payload.patchControlPoints)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.viewportCount, sizeof(payload.viewportCount)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.polygonMode, sizeof(payload.polygonMode)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.cullMode, sizeof(payload.cullMode)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.frontFace, sizeof(payload.frontFace)));
+        XXHASH_VERIFY(
+            XXH64_update(m_hashState, &payload.provokingVertexMode, sizeof(payload.provokingVertexMode)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.depthTestEnable, sizeof(payload.depthTestEnable)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.depthWriteEnable, sizeof(payload.depthWriteEnable)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.depthBiasEnable, sizeof(payload.depthBiasEnable)));
@@ -250,6 +253,23 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
 
         VkPipeline pipeline = CreatePipeline(payload);
+        // A failed creation must never be memoized. Caching VK_NULL_HANDLE served the null back for
+        // the rest of the process, so one transient driver rejection turned every later draw with
+        // the same state into a vkCmdBindPipeline(VK_NULL_HANDLE) - the SIGSEGV behind 9 of the 15
+        // CTS process deaths. Retrying costs one failed vkCreateGraphicsPipelines per draw, which
+        // is the correct price for a broken pipeline and is bounded by the draw itself being
+        // skipped.
+        if (pipeline == VK_NULL_HANDLE) {
+            // Unlatched, like the CreatePipeline report it accompanies: a pipeline MobileGL
+            // assembled and the driver refused is a broken invariant, not an expected failure,
+            // so it stays loud for as long as it is reachable. Raised from MGLOG_I once the
+            // Log.h ordering fix made MGLOG_E live in INFO builds.
+            MGLOG_E("PipelineFactory::GetOrCreatePipeline: creation failed for hash=0x%llx "
+                    "programHash=0x%llx; not caching the failure",
+                    static_cast<unsigned long long>(hash),
+                    static_cast<unsigned long long>(payload.programHash));
+            return VK_NULL_HANDLE;
+        }
         m_cache.emplace(hash, PipelineCacheEntry{pipeline, payload.programHash, payload.renderPass,
                                                  m_frameCounter});
         return pipeline;
@@ -387,8 +407,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         tessellation.patchControlPoints = payload.patchControlPoints;
 
         VkPipelineViewportStateCreateInfo vpci{VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO};
-        vpci.viewportCount = 1;
+        // Both counts move together: GL has one scissor rectangle per viewport, and Vulkan
-        vpci.scissorCount = 1;
+        // requires viewportCount == scissorCount whenever both are dynamic
+        // (VUID-VkPipelineViewportStateCreateInfo-scissorCount-04136). The caller has already
+        // clamped this to the device's multiViewport capability.
+        vpci.viewportCount = std::max<Uint32>(payload.viewportCount, 1u);
+        vpci.scissorCount = vpci.viewportCount;
 
         VkPipelineRasterizationStateCreateInfo raster{VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO};
         raster.polygonMode = payload.polygonMode;
@@ -397,6 +421,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         raster.depthBiasEnable = payload.depthBiasEnable ? VK_TRUE : VK_FALSE;
         raster.rasterizerDiscardEnable = payload.rasterizerDiscardEnable ? VK_TRUE : VK_FALSE;
         raster.lineWidth = 1.0f;
+        // Only chain the struct when the mode is not Vulkan's implicit default: a device without
+        // VK_EXT_provoking_vertex enabled must never see this pNext entry, and the renderer's
+        // selector already collapses to FIRST in exactly that case - so a device without the
+        // extension produces a byte-identical VkGraphicsPipelineCreateInfo to before.
+        VkPipelineRasterizationProvokingVertexStateCreateInfoEXT provokingVertexState{
+            VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_PROVOKING_VERTEX_STATE_CREATE_INFO_EXT};
+        if (payload.provokingVertexMode != VK_PROVOKING_VERTEX_MODE_FIRST_VERTEX_EXT) {
+            provokingVertexState.provokingVertexMode = payload.provokingVertexMode;
+            provokingVertexState.pNext = raster.pNext;
+            raster.pNext = &provokingVertexState;
+        }
 
         VkPipelineMultisampleStateCreateInfo ms{VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO};
         ms.rasterizationSamples = payload.rasterizationSamples;
@@ -445,9 +480,56 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         blend.attachmentCount = payload.colorAttachmentCount;
         blend.pAttachments = colorAttachments.empty() ? nullptr : colorAttachments.data();
 
+        // A GL program may have a tessellation EVALUATION stage and no CONTROL stage: GL 4.6 core
+        // 11.2.2 gives it a fixed-function pass-through instead. Vulkan has no such stage, and
+        // VUID-VkGraphicsPipelineCreateInfo-pStages-00730 requires both tessellation stages or
+        // neither - so the renderer synthesizes the pass-through GL describes and hands it in
+        // here (see ProgramFactory::GetOrCreatePassthroughTessControlStage).
+        //
+        // The refusal below is what keeps the half-tessellated shape away from the driver when
+        // there is no synthesized stage to add - because Mali does not reject it, it dereferences
+        // null INSIDE vkCreateGraphicsPipelines and takes the process down (SIGSEGV, fault addr
+        // 0x34, on Mali-G715/r54p2 and Mali-G925/r49p1 alike; Adreno and lavapipe merely render
+        // wrong). Returning VK_NULL_HANDLE routes this through the same path a driver rejection
+        // takes: the draw is skipped, nothing is memoised, and the process survives.
+        const Vector<VkPipelineShaderStageCreateInfo>* effectiveStages = payload.stages;
+        Vector<VkPipelineShaderStageCreateInfo> stagesWithPassthrough;
+        if (payload.passthroughTessControlStage.module != VK_NULL_HANDLE) {
+            stagesWithPassthrough = *payload.stages;
+            stagesWithPassthrough.push_back(payload.passthroughTessControlStage);
+            effectiveStages = &stagesWithPassthrough;
+        }
+        {
+            VkShaderStageFlags stagesPresent = 0;
+            for (const auto& stageInfo : *effectiveStages) {
+                stagesPresent |= stageInfo.stage;
+            }
+            const Bool hasTessControl = (stagesPresent & VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT) != 0;
+            const Bool hasTessEval = (stagesPresent & VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT) != 0;
+            if (hasTessControl != hasTessEval) {
+                // Latched, and the latch is the point: a failed creation is deliberately never
+                // memoised (see GetOrCreatePipeline), so a program in this state re-enters here
+                // once per draw, every frame - and a refusal diagnostic that repeats per draw is
+                // noise, not a diagnostic. One line names the program; the draws it explains are
+                // all the same draw.
+                static Bool s_warnedHalfTessellatedPipeline = false;
+                if (!s_warnedHalfTessellatedPipeline) {
+                    s_warnedHalfTessellatedPipeline = true;
+                    MGLOG_E_ONCE("PipelineFactory::CreatePipeline: refusing a pipeline with %s tessellation stage and "
+                            "no %s stage (VUID-VkGraphicsPipelineCreateInfo-pStages-00730). programHash=0x%llx "
+                            "patchControlPoints=%u. Its draws are skipped; logged once.",
+                            hasTessEval ? "an evaluation" : "a control",
+                            hasTessEval ? "control" : "evaluation",
+                            static_cast<unsigned long long>(payload.programHash),
+                            payload.patchControlPoints);
+                }
+                return VK_NULL_HANDLE;
+            }
+        }
+
         VkGraphicsPipelineCreateInfo gpi{VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO};
-        gpi.stageCount = static_cast<Uint32>(payload.stages->size());
+        gpi.stageCount = static_cast<Uint32>(effectiveStages->size());
-        gpi.pStages = payload.stages->data();
+        gpi.pStages = effectiveStages->data();
         gpi.pVertexInputState = payload.vertexInputState;
         gpi.pInputAssemblyState = &ia;
         gpi.pTessellationState =
@@ -464,6 +546,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         VkPipeline pipeline = VK_NULL_HANDLE;
         const VkResult result = vkCreateGraphicsPipelines(m_device, m_pipelineCache, 1, &gpi, nullptr, &pipeline);
+        // Loud, at MGLOG_F, and deliberately NOT latched. vkCreateGraphicsPipelines refusing a
+        // pipeline MobileGL assembled is a should-never-happen state, and the driver's own
+        // answer is VK_ERROR_UNKNOWN - no information at all - so this dump is the entire
+        // diagnosis. It is not an expected failure mode, so the one-shot rule that quiets W/E
+        // does not apply: while this is reachable it should keep saying so on every draw.
+        // GetOrCreatePipeline deliberately does not cache the failure, which is what makes that
+        // repetition happen; if the repetition ever needs to stop, fix the pipeline, not the log.
         if (result != VK_SUCCESS) {
             MGLOG_F("PipelineFactory::CreatePipeline failed: result=%s (%d) programHash=0x%llx vertexInputHash=0x%llx stageCount=%u topology=%s(%d) colorAttachmentCount=%u samples=%s(%d) subpass=%u",
                     VkResultToString(result),
@@ -494,6 +583,36 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             MGLOG_F("PipelineFactory::CreatePipeline vertex input: bindingCount=%u attributeCount=%u",
                     payload.vertexInputState->vertexBindingDescriptionCount,
                     payload.vertexInputState->vertexAttributeDescriptionCount);
+            // The driver's own answer is VK_ERROR_UNKNOWN, i.e. no information at all, so the only
+            // way to work out WHICH shader it choked on (the open sampler-array-in-struct
+            // investigation) is to name the modules. MGLOG_I, not _D: this is part of a
+            // should-never-happen report and must survive in the INFO-level builds that CTS
+            // actually runs against, alongside the MGLOG_F lines above.
+            if (payload.stageSpirvDigests) {
+                for (SizeT i = 0; i < payload.stageSpirvDigests->size(); ++i) {
+                    const auto& digest = (*payload.stageSpirvDigests)[i];
+                    MGLOG_I("PipelineFactory::CreatePipeline spirv[%zu]: stage=0x%x words=%u bytes=%zu "
+                            "hash=0x%llx",
+                            i, digest.stage, digest.wordCount,
+                            static_cast<SizeT>(digest.wordCount) * sizeof(Uint32),
+                            static_cast<unsigned long long>(digest.hash));
+                }
+            } else {
+                MGLOG_I("PipelineFactory::CreatePipeline: no SPIR-V digests attached to the payload");
+            }
+            if (payload.stages) {
+                for (SizeT i = 0; i < payload.stages->size(); ++i) {
+                    const auto& stage = (*payload.stages)[i];
+                    // VkShaderModule is a non-dispatchable handle: a pointer on 64-bit but a
+                    // plain uint64_t on 32-bit ABIs, where a cast to const void* is ill-formed
+                    // (broke the armeabi-v7a build). Print it as the 64-bit value it is.
+                    MGLOG_I("PipelineFactory::CreatePipeline stage[%zu]: stage=0x%x module=0x%llx entry=%s "
+                            "specialization=%d",
+                            i, static_cast<Uint32>(stage.stage),
+                            static_cast<unsigned long long>(reinterpret_cast<Uint64>(stage.module)),
+                            stage.pName ? stage.pName : "(null)", stage.pSpecializationInfo ? 1 : 0);
+                }
+            }
             for (Uint32 i = 0; i < payload.colorAttachmentCount; ++i) {
                 const auto& attachment = payload.colorBlendAttachments[i];
                 MGLOG_F("PipelineFactory::CreatePipeline colorAttachment[%u]: blend=%d colorWriteMask=0x%x srcColor=%d dstColor=%d colorOp=%d srcAlpha=%d dstAlpha=%d alphaOp=%d",

MobileGL/MG_Backend/DirectVulkan/Renderer/PipelineFactory.h

@@ -14,6 +14,16 @@
 #include <Includes.h>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
+    // Enough of a fingerprint to identify the exact module the driver rejected without keeping the
+    // SPIR-V alive for every program in the cache: a driver that answers VK_ERROR_UNKNOWN tells us
+    // nothing, so the log has to carry the shader's identity itself. Diagnostic only - never part
+    // of any pipeline or program hash.
+    struct ShaderStageSpirvDigest {
+        Uint32 stage = 0; // VkShaderStageFlagBits
+        Uint32 wordCount = 0;
+        Uint64 hash = 0;
+    };
+
     class PipelineFactory {
     public:
         using HashType = Uint64;
@@ -32,9 +42,22 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Bool primitiveRestartEnable = false;
             // GL_PATCH_VERTICES; only read for a PATCH_LIST topology.
             Uint32 patchControlPoints = 3;
+            // How many of ARB_viewport_array's viewports this pipeline rasterizes into. 1 for
+            // every program that never assigns gl_ViewportIndex, which is all of them outside the
+            // conformance suite - the wide shape costs a longer vkCmdSetViewport/Scissor per state
+            // change and can cost hardware fast paths, so it is opt-in per program. Baked into the
+            // pipeline (viewportCount is not dynamic without VK_EXT_extended_dynamic_state) and
+            // therefore hashed; the DYNAMIC viewport/scissor arrays the draw pushes must have
+            // exactly this many elements (VUID-vkCmdDraw-viewportCount-03417/-03418).
+            Uint32 viewportCount = 1;
             VkPolygonMode polygonMode = VK_POLYGON_MODE_FILL;
             VkCullModeFlags cullMode = VK_CULL_MODE_BACK_BIT;
             VkFrontFace frontFace = VK_FRONT_FACE_CLOCKWISE;
+            // GL's provoking vertex, baked into the pipeline (VK_EXT_provoking_vertex). It selects
+            // which vertex a flat varying takes AND the vertex order transform feedback records for
+            // strips/fans, so it is part of the pipeline's identity, not dynamic state. Defaults to
+            // Vulkan's own convention, which is what a device without the extension gets.
+            VkProvokingVertexModeEXT provokingVertexMode = VK_PROVOKING_VERTEX_MODE_FIRST_VERTEX_EXT;
             Bool depthTestEnable = false;
             Bool depthWriteEnable = false;
             Bool depthBiasEnable = false;
@@ -56,7 +79,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Bool fragmentReplacesDepth = false;
             Array<VkPipelineColorBlendAttachmentState, kMaxColorAttachments> colorBlendAttachments{};
             const Vector<VkPipelineShaderStageCreateInfo>* stages = nullptr;
+            // The tessellation control stage this renderer synthesized for a program that has
+            // an evaluation stage and none of its own (GL 4.6 core 11.2.2 gives such a program a
+            // fixed-function pass-through; Vulkan has no such thing and
+            // VUID-VkGraphicsPipelineCreateInfo-pStages-00730 forbids the half-tessellated
+            // pipeline outright). Appended to `stages` at creation. A null module means the
+            // renderer could not build one, and CreatePipeline refuses the pipeline - the same
+            // refusal it applies when `stages` itself is half-tessellated.
+            //
+            // NOT hashed: it is a pure function of the program and of patchControlPoints, both
+            // of which ComputeHash already mixes in.
+            VkPipelineShaderStageCreateInfo passthroughTessControlStage{};
             const VkPipelineVertexInputStateCreateInfo* vertexInputState = nullptr;
+            // Diagnostic only; may be null. Read solely from the pipeline-creation failure path.
+            const Vector<ShaderStageSpirvDigest>* stageSpirvDigests = nullptr;
         };
 
         explicit PipelineFactory(VkDevice device, const VulkanRendererConfig& config);

MobileGL/MG_Backend/DirectVulkan/Renderer/ProgramFactory.h

@@ -9,6 +9,7 @@
 #pragma once
 
 #include "../VkIncludes.h"
+#include "PipelineFactory.h"
 #include "MG_State/GLState/ProgramState/ProgramObject.h"
 #include "MG_State/GLState/ProgramState/ShaderObject.h"
 #include "MG_State/GLState/TextureState/TextureEnum.h"
@@ -32,7 +33,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             CombinedImageSampler,
             UniformTexelBuffer,
             StorageBuffer,
-            StorageImage
+            StorageImage,
+            // GLSL `imageBuffer` - a buffer texture reached through an IMAGE unit rather than a
+            // texture unit. Vulkan spells it VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER, which is a
+            // VkBufferView like UniformTexelBuffer and not a VkImageView like StorageImage: it is
+            // the one image uniform whose descriptor is a buffer. Appended, never inserted -
+            // DescriptorKeyHash mixes the enumerator's value.
+            StorageTexelBuffer
         };
 
         enum class CompileOptionBit : Uint {
@@ -52,25 +59,69 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             // recorded while GL transform feedback is active, so plain draws keep the
             // undecorated variant.
             XfbCapture = 1 << 6,
+            // Rewrites the fragment stage's gl_FragCoord reads to GL's bottom-left window
+            // origin. Vulkan's gl_FragCoord.y IS the framebuffer row being written, and the
+            // default framebuffer's image is stored in display (top-left) order, so a shader
+            // that reads gl_FragCoord there sees `height - y_GL`. Set together with
+            // PositionYFlip (the two are the same fact about the same draws) except under a
+            // quarter turn, which this renderer does not convert rectangles for either.
+            FragCoordYFlip = 1 << 7,
+            // Replaces the vertex stage's gl_BaseVertex reads with zero. GL defines the builtin
+            // as zero for every drawing command that has no baseVertex parameter - all the
+            // DrawArrays forms - while Vulkan's BaseVertex reports firstVertex there. Set only
+            // for a non-indexed draw whose program actually reads the builtin, so nothing else
+            // acquires a second program/pipeline variant. See ZeroBaseVertexPass.
+            ZeroBaseVertex = 1 << 8,
         };
         using CompileOptionFlags = Flags<CompileOptionBit>;
         using HashType = Uint64;
 
+        struct UpdateAfterBindLimits {
+            Bool enabled = false;
+            Uint32 maxPerStageSamplers = 0;
+            Uint32 maxPerStageUniformBuffers = 0;
+            Uint32 maxPerStageStorageBuffers = 0;
+            Uint32 maxPerStageSampledImages = 0;
+            Uint32 maxPerStageStorageImages = 0;
+            Uint32 maxPerStageResources = 0;
+            Uint32 maxSetSamplers = 0;
+            Uint32 maxSetUniformBuffers = 0;
+            Uint32 maxSetUniformBuffersDynamic = 0;
+            Uint32 maxSetStorageBuffers = 0;
+            Uint32 maxSetStorageBuffersDynamic = 0;
+            Uint32 maxSetSampledImages = 0;
+            Uint32 maxSetStorageImages = 0;
+        };
+
         struct VkProgramObject {
             static constexpr Uint32 kMaxVertexInputLocations = 32;
 
             HashType hash = 0;
             Vector<VkPipelineShaderStageCreateInfo> stages;
             Vector<VkShaderModule> modules;
+            // Parallel to stages; identifies the exact module bytes handed to the driver when a
+            // pipeline creation fails. Sixteen bytes per stage instead of keeping the SPIR-V.
+            Vector<ShaderStageSpirvDigest> stageSpirvDigests;
 
             // Layout data (previously in separate VkProgramLayout)
             VkDescriptorSetLayout descriptorSetLayout = VK_NULL_HANDLE;
+            // True only when this layout passed every descriptor-indexing feature and
+            // update-after-bind limit gate at reflection time. It controls both the
+            // layout/binding flags and the pool class used by UniformManager.
+            Bool usesUpdateAfterBind = false;
             VkPipelineLayout pipelineLayout = VK_NULL_HANDLE;
             Vector<DescriptorBindingKind> bindingKinds;
+            // The bindings this program actually declares, ascending. bindingKinds is sized to the
+            // 256-binding cap while a real GL program uses 1-8, so the per-draw descriptor walk was
+            // scanning 256 slots to find a handful. MUST stay ascending: Vulkan consumes
+            // pDynamicOffsets in binding order and the writer pushes them in iteration order, so an
+            // unordered list would silently mis-pair dynamic offsets with their uniform blocks.
+            Vector<Uint32> activeBindings;
             Vector<Uint32> dynamicBindings;
             Vector<Int> uniformBlockIndexByBinding;
-            // Descriptor count per binding (1 except for UBO instance arrays, which occupy one
+            // Descriptor count per binding (1 except for a descriptor ARRAY - a UBO or storage
-            // binding with descriptorCount = N).
+            // block instance array, an image uniform array or a sampler uniform array - each of
+            // which occupies one binding with descriptorCount = N).
             Vector<Uint16> bindingDescriptorCounts;
             // Per-element GL uniform block indices for arrayed UBO bindings (count > 1);
             // element 0 of a non-arrayed binding stays in uniformBlockIndexByBinding.
@@ -79,6 +130,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Vector<Int> samplerUniformLocationByBinding;
             Vector<TextureTarget> samplerTextureTargetByBinding;
             Vector<SamplerNumericDomain> samplerNumericDomainByBinding;
+            // Shared by StorageImage and StorageTexelBuffer bindings: a binding is one kind or
+            // the other, never both, and both need exactly the same thing - the format the
+            // shader declared, so the per-draw resolve can tell a typed declaration from a
+            // formatless one. Kept as one pair rather than two so the move operations below
+            // cannot drift out of sync with a field that only one kind populates.
             Vector<VkFormat> storageImageFormatByBinding;
             Vector<Bool> storageImageUsesBindingFormatByBinding;
             Vector<String> storageBlockNameByBinding;
@@ -86,6 +142,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             // Set once during ReflectLayout so the per-draw path can skip the whole
             // storage-image preparation for the overwhelming majority of programs.
             Bool hasStorageImages = false;
+            // Something about this program's descriptors could not be resolved - an opaque
+            // uniform array whose elements have no addressable uniform locations (the
+            // multi-dimensional case), or a binding remap that failed outright. The binding
+            // STAYS DECLARED in the descriptor set layout; declining is done here, by refusing
+            // every draw, and BindProgramUniformBuffers returns false so the draw setup skips
+            // the draw exactly as it does for any other bind failure.
+            //
+            // Keeping the layout intact is the load-bearing half. Shrinking it instead - which
+            // is what the first cut of this did - leaves the shader reading a descriptor the
+            // layout never declared, and lavapipe segfaults on that inside PIPELINE CREATION,
+            // in a JIT worker thread, before any draw runs where a refusal could help. The
+            // reason was logged once at MGLOG_I when the descriptor was declined.
+            Bool declinedDescriptors = false;
             Int globalUboBinding = -1;
             Uint32 activeVertexInputLocationMask = 0;
             Array<GLenum, kMaxVertexInputLocations> vertexInputTypes{};
@@ -98,9 +167,37 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             // gl_FragDepth); shader-computed depth is immune to the cross-pipeline
             // position-invariance quirk (see PipelineFactory::ShouldSuppressDepthWrite).
             Bool fragmentReplacesDepth = false;
+            // The vertex module declares the BaseVertex builtin. Selects the ZeroBaseVertex
+            // program variant for non-indexed draws, and is deliberately a property of the
+            // PROGRAM rather than of the variant: the zeroed variant leaves the variable
+            // declared, so both variants answer the same and the draw path can ask either.
+            Bool readsBaseVertexBuiltin = false;
+            // Some pre-rasterization stage assigns gl_ViewportIndex. Its pipeline declares
+            // viewportCount = the renderer's rasterizable viewport count instead of 1, and its
+            // draws push the whole viewport/scissor array; every other program keeps the
+            // single-viewport fast path untouched. Part of the program's identity (folded into
+            // the pipeline hash through programHash), so no memo can serve the wrong shape.
+            Bool writesViewportIndexBuiltin = false;
+            // This program has a tessellation EVALUATION stage and no tessellation CONTROL
+            // stage. GL allows that (4.6 core 11.2.2: with no control shader the input patch
+            // is passed through unmodified, the output patch size is PATCH_VERTICES, and the
+            // levels come from the PATCH_DEFAULT_*_LEVEL state); Vulkan does not - either both
+            // tessellation stages are present or neither
+            // (VUID-VkGraphicsPipelineCreateInfo-pStages-00730). So the draw path has to supply
+            // the pass-through stage GL describes; see GetOrCreatePassthroughTessControlStage.
+            Bool needsPassthroughTessControl = false;
+            // ...and the pass-through this renderer can synthesize carries gl_Position and
+            // nothing else, so it is only correct when the evaluation stage's inputs are
+            // built-ins. A user-defined varying would arrive at the evaluation stage
+            // UNWRITTEN once a control stage sits between it and the vertex stage, which is
+            // silently wrong pixels rather than a crash - so those programs are declined
+            // instead (PipelineFactory::CreatePipeline refuses the pipeline and the draw is
+            // skipped). See ReflectPassthroughTessControlNeed.
+            Bool passthroughTessControlEmulatable = false;
             // Frame-boundary counter value of the last GetOrCreateProgram hit; drives
-            // cache eviction (see OnFrameBoundary).
+            // cache eviction (see OnFrameBoundary). Mutable: the draw snapshot's memoised
-            Uint64 lastUsedFrame = 0;
+            // entry pointer re-stamps use through a const reference (StampProgramUse).
+            mutable Uint64 lastUsedFrame = 0;
 
             static inline VkDevice s_device = VK_NULL_HANDLE;
 
@@ -111,9 +208,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 hash = other.hash;
                 stages = std::move(other.stages);
                 modules = std::move(other.modules);
+                // Must travel with `modules`: these digests name the SPIR-V those exact
+                // shader modules were built from, and the pipeline-failure diagnostics
+                // print the two together. Leaving it behind used to merely lose the
+                // digests on a rehash; now that the cache is a robin-hood table, insertion
+                // SWAPS two entries, and a field that no move touches stays behind in the
+                // slot - pairing one program's modules with another program's digests, so
+                // a pipeline failure would be reported against the wrong SPIR-V.
+                stageSpirvDigests = std::move(other.stageSpirvDigests);
                 descriptorSetLayout = other.descriptorSetLayout;
+                usesUpdateAfterBind = other.usesUpdateAfterBind;
                 pipelineLayout = other.pipelineLayout;
                 bindingKinds = std::move(other.bindingKinds);
+                activeBindings = std::move(other.activeBindings);
                 dynamicBindings = std::move(other.dynamicBindings);
                 uniformBlockIndexByBinding = std::move(other.uniformBlockIndexByBinding);
                 bindingDescriptorCounts = std::move(other.bindingDescriptorCounts);
@@ -128,6 +235,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 storageBlockNameByBinding = std::move(other.storageBlockNameByBinding);
                 storageBlockIndexByBinding = std::move(other.storageBlockIndexByBinding);
                 hasStorageImages = other.hasStorageImages;
+                declinedDescriptors = other.declinedDescriptors;
                 globalUboBinding = other.globalUboBinding;
                 activeVertexInputLocationMask = other.activeVertexInputLocationMask;
                 vertexInputTypes = other.vertexInputTypes;
@@ -137,11 +245,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 producerOutputComponentCount = other.producerOutputComponentCount;
                 fragmentInputComponentCount = other.fragmentInputComponentCount;
                 fragmentReplacesDepth = other.fragmentReplacesDepth;
+                readsBaseVertexBuiltin = other.readsBaseVertexBuiltin;
+                writesViewportIndexBuiltin = other.writesViewportIndexBuiltin;
+                needsPassthroughTessControl = other.needsPassthroughTessControl;
+                passthroughTessControlEmulatable = other.passthroughTessControlEmulatable;
                 lastUsedFrame = other.lastUsedFrame;
                 other.hash = 0;
                 other.descriptorSetLayout = VK_NULL_HANDLE;
+                other.usesUpdateAfterBind = false;
                 other.pipelineLayout = VK_NULL_HANDLE;
                 other.hasStorageImages = false;
+                other.declinedDescriptors = false;
                 other.globalUboBinding = -1;
                 other.activeVertexInputLocationMask = 0;
                 other.activeFragmentOutputLocationMask = 0;
@@ -149,6 +263,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 other.producerOutputComponentCount = 0;
                 other.fragmentInputComponentCount = 0;
                 other.fragmentReplacesDepth = false;
+                other.readsBaseVertexBuiltin = false;
+                other.writesViewportIndexBuiltin = false;
+                other.needsPassthroughTessControl = false;
+                other.passthroughTessControlEmulatable = false;
                 other.lastUsedFrame = 0;
             }
             VkProgramObject& operator=(VkProgramObject&& other) noexcept {
@@ -159,9 +277,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 hash = other.hash;
                 stages = std::move(other.stages);
                 modules = std::move(other.modules);
+                stageSpirvDigests = std::move(other.stageSpirvDigests); // travels with `modules` - see the move ctor
                 descriptorSetLayout = other.descriptorSetLayout;
+                usesUpdateAfterBind = other.usesUpdateAfterBind;
                 pipelineLayout = other.pipelineLayout;
                 bindingKinds = std::move(other.bindingKinds);
+                activeBindings = std::move(other.activeBindings);
                 dynamicBindings = std::move(other.dynamicBindings);
                 uniformBlockIndexByBinding = std::move(other.uniformBlockIndexByBinding);
                 bindingDescriptorCounts = std::move(other.bindingDescriptorCounts);
@@ -176,6 +297,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 storageBlockNameByBinding = std::move(other.storageBlockNameByBinding);
                 storageBlockIndexByBinding = std::move(other.storageBlockIndexByBinding);
                 hasStorageImages = other.hasStorageImages;
+                declinedDescriptors = other.declinedDescriptors;
                 globalUboBinding = other.globalUboBinding;
                 activeVertexInputLocationMask = other.activeVertexInputLocationMask;
                 vertexInputTypes = other.vertexInputTypes;
@@ -185,11 +307,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 producerOutputComponentCount = other.producerOutputComponentCount;
                 fragmentInputComponentCount = other.fragmentInputComponentCount;
                 fragmentReplacesDepth = other.fragmentReplacesDepth;
+                readsBaseVertexBuiltin = other.readsBaseVertexBuiltin;
+                writesViewportIndexBuiltin = other.writesViewportIndexBuiltin;
+                needsPassthroughTessControl = other.needsPassthroughTessControl;
+                passthroughTessControlEmulatable = other.passthroughTessControlEmulatable;
                 lastUsedFrame = other.lastUsedFrame;
                 other.hash = 0;
                 other.descriptorSetLayout = VK_NULL_HANDLE;
+                other.usesUpdateAfterBind = false;
                 other.pipelineLayout = VK_NULL_HANDLE;
                 other.hasStorageImages = false;
+                other.declinedDescriptors = false;
                 other.globalUboBinding = -1;
                 other.activeVertexInputLocationMask = 0;
                 other.activeFragmentOutputLocationMask = 0;
@@ -197,6 +325,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 other.producerOutputComponentCount = 0;
                 other.fragmentInputComponentCount = 0;
                 other.fragmentReplacesDepth = false;
+                other.readsBaseVertexBuiltin = false;
+                other.writesViewportIndexBuiltin = false;
+                other.needsPassthroughTessControl = false;
+                other.passthroughTessControlEmulatable = false;
                 other.lastUsedFrame = 0;
                 return *this;
             }
@@ -224,6 +356,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 }
                 modules.clear();
                 stages.clear();
+                stageSpirvDigests.clear(); // the modules they describe are gone
             }
         };
 
@@ -239,21 +372,72 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             virtual void OnProgramEvicted(HashType programHash, VkDescriptorSetLayout descriptorSetLayout) = 0;
         };
 
-        explicit ProgramFactory(VkDevice device, const VulkanRendererConfig& config, Uint32 maxBindings = 16,
+        // How this factory's compute modules implement GL_KHR_shader_subgroup. Computed
-                                Bool shaderDrawParametersEnabled = false,
+        // once at renderer initialization (SubgroupSupportPolicy.h + the device's
-                                Bool unformattedFloatStorageImagesEnabled = false)
+        // subgroup properties) so lowering can never disagree with the advertised
+        // capabilities. Native subgroup operations always execute natively; the two
+        // repair passes patch modules AROUND them, and the emulation only replaces them
+        // on opted-in devices with no subgroup support at all.
+        struct SubgroupLoweringPolicy {
+            Bool emulateSubgroups = false;      // MOBILEGL_MAGMA_EMULATE_SUBGROUP, no-native-support devices
+            Bool fixIterationRPSubgroupScratch = false; // patch iterationRP's under-declared scratch
+            Bool fixIterationRPBarrier = false; // repair Program 203's shared-scratch race
+            Bool deriveNumSubgroups = false;    // repair the NumSubgroups builtin
+            Bool requireFullSubgroups = false;  // computeFullSubgroups enabled on the device
+            Uint32 nativeSubgroupSize = 0;
+            // Full-subgroup launches are bounded by this device limit; a dispatch whose
+            // workgroup needs more subgroups than this cannot request the flag.
+            Uint32 maxComputeWorkgroupSubgroups = 0;
+            // VkPhysicalDeviceLimits::maxComputeSharedMemorySize; bounds the scratch the
+            // emulation pass may add (0 falls back to the Vulkan minimum, 16384).
+            Uint32 maxComputeSharedMemoryBytes = 0;
+        };
+
+        explicit ProgramFactory(VkDevice device, const VulkanRendererConfig& config, Uint32 maxBindings,
+                                Bool shaderDrawParametersEnabled,
+                                Bool unformattedFloatStorageImagesEnabled,
+                                Bool enableSpirvValidation,
+                                UpdateAfterBindLimits updateAfterBindLimits,
+                                SubgroupLoweringPolicy subgroupPolicy)
             : m_device(device), m_maxBindings(maxBindings), m_config(config),
               m_shaderDrawParametersEnabled(shaderDrawParametersEnabled),
-              m_unformattedFloatStorageImagesEnabled(unformattedFloatStorageImagesEnabled) {
+              m_unformattedFloatStorageImagesEnabled(unformattedFloatStorageImagesEnabled),
+              m_enableSpirvValidation(enableSpirvValidation),
+              m_updateAfterBindLimits(updateAfterBindLimits),
+              m_subgroupPolicy(subgroupPolicy) {
             VkProgramObject::s_device = device;
         }
-        ~ProgramFactory() = default;
+        // Destroys the pass-through tessellation control modules. Runs while the device is
+        // still alive for the same reason ~VkProgramObject's does: this factory outlives
+        // nothing that owns the device.
+        ~ProgramFactory();
         ProgramFactory(const ProgramFactory&) = delete;
 
         HashType ComputeHash(const MG_State::GLState::ProgramObject& program, CompileOptionFlags flags) const;
         const VkProgramObject& GetOrCreateProgram(
             const MG_State::GLState::ProgramObject& program, CompileOptionFlags flags);
 
+        // The default framebuffer's current image height, baked as a literal into every
+        // FragCoordYFlip variant (there is no push-constant or specialization channel here, and
+        // adding one for a value that changes only on swapchain recreation would cost the draw
+        // path more than a recompile costs a resize). It is therefore part of those variants'
+        // identity: ComputeHash mixes it in when the bit is set, so a height change re-keys them
+        // and leaves every other program's hash untouched. Setting a NEW height also bumps the
+        // cache-structure epoch, because a caller holding a memoised VkProgramObject* would
+        // otherwise keep using a module compiled against the old height.
+        void SetDefaultFramebufferHeight(Uint32 height);
+        Uint32 GetDefaultFramebufferHeight() const { return m_defaultFramebufferHeight; }
+
+        // Bumped whenever m_cache's STRUCTURE changes (any insert or erase): the cache is
+        // an open-addressing map holding entries by value, so both moves existing entries.
+        // A caller that memoised a VkProgramObject* may keep dereferencing it only while
+        // this is unchanged; on a bump it must re-run GetOrCreateProgram.
+        Uint64 GetCacheStructureEpoch() const { return m_cacheStructureEpoch; }
+        // A memoised entry pointer bypasses GetOrCreateProgram, whose per-lookup stamp is
+        // what keeps an in-use entry out of OnFrameBoundary's idle sweep - so such a
+        // caller must re-stamp the entry itself, at least once per frame boundary.
+        void StampProgramUse(const VkProgramObject& entry) const { entry.lastUsedFrame = m_frameCounter; }
+
         // Observer may be null (no notifications). Not owned.
         void SetEvictionObserver(IEvictionObserver* observer) { m_evictionObserver = observer; }
         // Frame boundary hook: ages the program cache and evicts long-unused entries
@@ -272,6 +456,39 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // True when an entry point reads the InstanceIndex builtin. Only gates a diagnostic:
         // without shaderDrawParameters such a shader cannot have gl_InstanceID rebased.
         static Bool ReflectedReadsInstanceIndexBuiltin(const SpvReflectShaderModule& reflectModule);
+        // True when an entry point declares the BaseVertex builtin, i.e. when a non-indexed
+        // draw with this program has to take the ZeroBaseVertex variant.
+        static Bool ReflectedReadsBaseVertexBuiltin(const SpvReflectShaderModule& reflectModule);
+        // Shared by the two above: does any entry point list an input variable decorated with
+        // this builtin?
+        static Bool ReflectedDeclaresInputBuiltin(const SpvReflectShaderModule& reflectModule, SpvBuiltIn builtin);
+        // True when an entry point writes the ViewportIndex builtin (gl_ViewportIndex), i.e. when
+        // the program can route primitives to a viewport other than 0 and its pipeline therefore
+        // has to declare more than one. Asks about OUTPUT variables because that is the direction
+        // a pre-rasterization stage declares it in.
+        static Bool ReflectedWritesViewportIndexBuiltin(const SpvReflectShaderModule& reflectModule);
+        static Bool ReflectedDeclaresOutputBuiltin(const SpvReflectShaderModule& reflectModule, SpvBuiltIn builtin);
+
+        // The pass-through tessellation control stage GL 4.6 core 11.2.2 describes for a
+        // program that has an evaluation stage and no control stage, for an input patch of
+        // `patchVertices` control points. Returned BY VALUE (a stage description is a POD, and
+        // the cache below is a rehashing map, so a pointer into it would not survive the next
+        // distinct patch size). `.module == VK_NULL_HANDLE` means the stage could not be built:
+        // the caller then has no control stage to inject, and CreatePipeline refuses the
+        // pipeline rather than handing the driver a half-tessellated one.
+        //
+        // Keyed on the patch size because GL takes the output patch size from PATCH_VERTICES,
+        // which is draw state, not link state - the CTS case that motivated this links at the
+        // default 3 and draws at 4. The pipeline cache already re-keys on patchControlPoints,
+        // so the module a pipeline was built with is part of that pipeline's identity.
+        // Compiling is bounded by the number of distinct patch sizes a program draws with
+        // (MAX_PATCH_VERTICES = 32 in the worst case, one or two in practice) and only ever
+        // happens for the rare program that has no control stage at all.
+        VkPipelineShaderStageCreateInfo GetOrCreatePassthroughTessControlStage(Uint32 patchVertices);
+
+        // Source of the module above. Exposed for tests: the generated GLSL is the whole
+        // contract with the evaluation stage, so it is worth pinning independently of a device.
+        static String BuildPassthroughTessControlSource(Uint32 patchVertices);
 
     private:
         struct ProgramLookupCache {
@@ -285,11 +502,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         void ReflectVertexInputs(const Vector<SharedPtr<MG_State::GLState::ShaderObject>>& shaders,
                      const Vector<Vector<Uint>>& spirv,
                      VkProgramObject& entry) const;
+        void ReflectViewportIndexUsage(const Vector<SharedPtr<MG_State::GLState::ShaderObject>>& shaders,
+                                       const Vector<Vector<Uint>>& spirv,
+                                       VkProgramObject& entry) const;
         void ReflectFragmentOutputs(const Vector<SharedPtr<MG_State::GLState::ShaderObject>>& shaders,
                         const Vector<Vector<Uint>>& spirv,
                         VkProgramObject& entry) const;
         void ReflectLayout(const MG_State::GLState::ProgramObject& program, const Vector<Vector<Uint>>& spirv,
                            VkProgramObject& entry) const;
+        // Fills needsPassthroughTessControl / passthroughTessControlEmulatable off the linked
+        // modules. Const and reflection-only: it decides nothing about the pipeline, it only
+        // records what the evaluation stage's input interface is made of.
+        void ReflectPassthroughTessControlNeed(const Vector<SharedPtr<MG_State::GLState::ShaderObject>>& shaders,
+                                               const Vector<Vector<Uint>>& spirv,
+                                               VkProgramObject& entry) const;
 
         VkDevice m_device = VK_NULL_HANDLE;
         Uint32 m_maxBindings = 0;
@@ -301,10 +527,28 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // True only when the logical device enabled both
         // shaderStorageImageReadWithoutFormat and shaderStorageImageWriteWithoutFormat.
         Bool m_unformattedFloatStorageImagesEnabled = false;
+        // Startup snapshot used only by internally synthesized shader modules, which do not
+        // originate from a ProgramLinkTask.
+        Bool m_enableSpirvValidation = false;
+        // Device feature and limit gate resolved before vkCreateDevice. Keeping it in
+        // the factory lets each reflected layout choose ordinary descriptors when its
+        // own counts would exceed the update-after-bind budget.
+        UpdateAfterBindLimits m_updateAfterBindLimits{};
+        SubgroupLoweringPolicy m_subgroupPolicy{};
+        // See SetDefaultFramebufferHeight. 0 means "not known yet"; the FragCoordYFlip bit is
+        // never set before the swapchain exists, so no variant can be compiled against it.
+        Uint32 m_defaultFramebufferHeight = 0;
         mutable ProgramLookupCache m_lastLookup;
         // Monotonic frame-boundary counter (bumped in OnFrameBoundary) for cache aging.
         Uint64 m_frameCounter = 0;
+        // See GetCacheStructureEpoch(). Starts at 1 so a zero-initialized memo can never match.
+        Uint64 m_cacheStructureEpoch = 1;
         IEvictionObserver* m_evictionObserver = nullptr;
+        // Pass-through tessellation control stages by input patch size. Never evicted: at most
+        // MAX_PATCH_VERTICES entries exist for the lifetime of the device, and every pipeline
+        // ever built from one keeps referencing its module. A failed build is cached as
+        // VK_NULL_HANDLE so a broken generator costs one compile, not one per draw.
+        UnorderedMap<Uint32, VkPipelineShaderStageCreateInfo> m_passthroughTessControlStages;
         static inline XXH64_state_t* m_hashState = XXH64_createState();
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/SwapchainObject.cpp

@@ -157,7 +157,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         MGLOG_I("Got %d surface formats:", swapchainCapabilities.surfaceFormats.size());
         for (const auto& sf : swapchainCapabilities.surfaceFormats) {
-            MGLOG_I("    [%s, %s]", string_VkFormat(sf.format), string_VkColorSpaceKHR(sf.colorSpace));
+            MGLOG_D("    [%s, %s]", string_VkFormat(sf.format), string_VkColorSpaceKHR(sf.colorSpace));
         }
 
         const auto pickedSurfaceFormat = ChooseSwapchainSurfaceFormat(swapchainCapabilities.surfaceFormats);
@@ -166,7 +166,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         MGLOG_I("Got %d present modes:", swapchainCapabilities.presentModes.size());
         for (const auto& pm : swapchainCapabilities.presentModes) {
-            MGLOG_I("    %s", string_VkPresentModeKHR(pm));
+            MGLOG_D("    %s", string_VkPresentModeKHR(pm));
         }
 
         const auto presentMode = ChooseSwapchainPresentMode(swapchainCapabilities.presentModes);

MobileGL/MG_Backend/DirectVulkan/Renderer/UniformManager.h

@@ -26,9 +26,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     public:
         struct SamplerBindingOverride {
             Uint32 binding = 0;
+            Uint32 element = 0;
             MG_State::GLState::ITextureObject* texture = nullptr;
             const MG_State::GLState::SamplerObject* sampler = nullptr;
             VkImageView imageView = VK_NULL_HANDLE;
+            VkImageLayout imageLayout = VK_IMAGE_LAYOUT_UNDEFINED;
+            Bool forceNearestFiltering = false;
+        };
+
+        struct SamplerImageFeedbackBinding {
+            Uint32 samplerBinding = 0;
+            Uint32 samplerElement = 0;
+            MG_State::GLState::ITextureObject* texture = nullptr;
+            const MG_State::GLState::SamplerObject* sampler = nullptr;
+            SamplerNumericDomain numericDomain = SamplerNumericDomain::Unknown;
         };
 
         Bool Initialize(VkDevice device, VkBufferManager* bufferManager,
@@ -53,18 +64,52 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // caches - a live layout's entry must never be purged (its sets would be
         // unreachable pool slots), so there is deliberately no age-based sweep here.
         void OnDescriptorSetLayoutDestroyed(VkDescriptorSetLayout descriptorSetLayout);
+        // One record per visited CombinedImageSampler DESCRIPTOR (post fallback substitution,
+        // in binding order, and within a binding in array-element order): the resolved texture
+        // and effective sampler, as never-reused lifetime ids so a freed-and-reallocated object
+        // at the same heap address can only MISS a comparison, never false-hit it (same ABA
+        // rule as SamplerResolveMemo). An arrayed binding contributes one record per element -
+        // element granularity is required, or swapping the textures of two elements of the same
+        // array would leave the record list identical and the fast path would keep a stale set.
+        struct SampledBindingRecord {
+            Uint64 textureLifetimeId = 0;
+            Uint64 samplerLifetimeId = 0;
+        };
         Bool CollectSampledTextures(const MG_State::GLState::ProgramObject& program,
                                     const ProgramFactory::VkProgramObject& programObj,
-                                    Vector<MG_State::GLState::ITextureObject*>& outTextures);
+                                    Vector<MG_State::GLState::ITextureObject*>& outTextures,
+                                    Vector<SampledBindingRecord>* outBindingRecords = nullptr);
+        // Shadow-compare for the SetupDraw fast path: re-runs the CollectSampledTextures
+        // walk and reports whether every visited binding still resolves to the recorded
+        // (texture, effective sampler) pair. A texture bind generation bump alone (e.g. a
+        // redundant glBindSampler, which always bumps it) does not prove the sampled set
+        // moved; this walk does, without rebuilding the set or falling off the fast path.
+        Bool SampledBindingsUnchanged(const MG_State::GLState::ProgramObject& program,
+                                      const ProgramFactory::VkProgramObject& programObj,
+                                      const Vector<SampledBindingRecord>& previousRecords) const;
         Bool CollectStorageImageTextures(const MG_State::GLState::ProgramObject& program,
                                          const ProgramFactory::VkProgramObject& programObj,
                                          Vector<MG_State::GLState::ITextureObject*>& outTextures) const;
+        Bool CollectSamplerImageFeedback(
+            const MG_State::GLState::ProgramObject& program,
+            const ProgramFactory::VkProgramObject& programObj,
+            Vector<SamplerImageFeedbackBinding>& outBindings) const;
+        static Bool SamplerOverlapsWritableImageSubresource(Int samplerBaseLevel, Int samplerMaxLevel,
+                                                             GLint imageLevel, GLenum imageAccess);
+        // samplerDescriptorsUnchangedHint: the caller (SetupDraw fast path) proved that
+        // every input of every combined-image-sampler resolution is unchanged since the
+        // previous draw's resolve - same (texture, sampler) per binding, texture params
+        // sum, sampling-resolution generation (sampler params + texture shape), image
+        // epochs AND per-resource layout values - so the per-binding cached
+        // VkDescriptorImageInfo may be reused without re-running the resolve chain.
         Bool BindProgramUniformBuffers(VkCommandBuffer commandBuffer,
                                        const MG_State::GLState::ProgramObject& program,
                                        const ProgramFactory::VkProgramObject& programObj,
                                        Uint32 frameIndex,
                                        VkPipelineBindPoint bindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
-                                       const SamplerBindingOverride* samplerBindingOverride = nullptr);
+                                       const SamplerBindingOverride* samplerBindingOverride = nullptr,
+                                       Bool samplerDescriptorsUnchangedHint = false,
+                                       const Vector<SamplerBindingOverride>* samplerBindingOverrides = nullptr);
 
         // Pure format-policy helper kept public for host regression tests. Formatted storage
         // images use their shader qualifier; transformed float images use glBindImageTexture's
@@ -87,6 +132,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VkDescriptorPool handle = VK_NULL_HANDLE;
             Uint32 maxSets = 0;
             Uint32 allocatedSets = 0;
+            Bool updateAfterBind = false;
         };
 
         // A cached descriptor set together with the pool it was allocated from, so a
@@ -114,28 +160,60 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         static Bool ResolveSamplerTexture(const MG_State::GLState::ProgramObject& program,
                                    const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
                                    SharedPtr<MG_State::GLState::ITextureObject>& outTexture);
+        // Shared per-binding resolution for CollectSampledTextures and
+        // SampledBindingsUnchanged, so membership and comparison can never diverge:
+        // texture after the fallback substitution (may still be null when no fallback
+        // exists), effective sampler = unit override else the texture's own sampler.
+        // False = the binding is skipped (unbound with a non-2D fallback target).
+        // `element` indexes a sampler array inside the binding; see ResolveSamplerDescriptor.
+        Bool ResolveSampledBinding(const MG_State::GLState::ProgramObject& program,
+                                   const ProgramFactory::VkProgramObject& programObj, Uint32 binding, Uint32 element,
+                                   MG_State::GLState::ITextureObject*& outTexture,
+                                   const MG_State::GLState::SamplerObject*& outSampler) const;
         // Raw-pointer variant for the per-draw sampled-texture walk (CollectSampledTextures):
         // the bound texture stays alive through the draw via GL binding state, so callers that
         // only need the pointer skip the SharedPtr copy's atomic refcount churn.
         static MG_State::GLState::ITextureObject* ResolveSamplerTextureRaw(
             const MG_State::GLState::ProgramObject& program,
-            const ProgramFactory::VkProgramObject& programObj, Uint32 binding);
+            const ProgramFactory::VkProgramObject& programObj, Uint32 binding, Uint32 element);
         SharedPtr<MG_State::GLState::ITextureObject> GetFallbackTexture(TextureTarget target) const;
+        // `element` indexes a sampler ARRAY inside one binding; each element carries its own
+        // independently assigned GL texture unit, so it selects the texture, the sampler
+        // override and the fallback separately from its neighbours.
+        //
+        // trustUnchangedHint: reuse this binding's cached VkDescriptorImageInfo outright
+        // (see BindProgramUniformBuffers' samplerDescriptorsUnchangedHint for the proof
+        // obligations the caller carries). The cache is keyed by binding alone, so it is
+        // used ONLY for single-descriptor bindings - see m_samplerResolveMemo.
         Bool ResolveSamplerDescriptor(VkCommandBuffer commandBuffer, const MG_State::GLState::ProgramObject& program,
                                       const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
-                                      VkDescriptorImageInfo& outImageInfo) const;
+                                      Uint32 element, VkDescriptorImageInfo& outImageInfo,
+                                      Bool trustUnchangedHint = false) const;
         Bool ResolveSamplerDescriptorOverride(const SamplerBindingOverride& samplerBindingOverride,
                                               VkDescriptorImageInfo& outImageInfo) const;
         Bool ResolveTexelBufferDescriptor(const MG_State::GLState::ProgramObject& program,
                                           const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
                                           Uint32 frameIndex, VkBufferView& outBufferView);
+        // GLSL `imageBuffer`: the same VkBufferView descriptor as the sampled texel buffer above,
+        // but resolved from an IMAGE unit (glBindImageTexture) rather than a texture unit, and
+        // made GPU-resident-writable because the shader may store to it. No `element` parameter:
+        // an imageBuffer ARRAY is refused at program creation, so a binding is always one
+        // descriptor (see the array gate in RemapDescriptorBindingsForVulkan).
+        Bool ResolveStorageTexelBufferDescriptor(const MG_State::GLState::ProgramObject& program,
+                                                 const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
+                                                 Uint32 frameIndex, VkBufferView& outBufferView);
+        // `element` indexes a block INSTANCE array's descriptors; it is 0 for every ordinary
+        // block. Each element resolves through its own GL storage block, and so its own GL
+        // binding point, buffer and glBindBufferRange window.
         Bool ResolveStorageBufferDescriptor(const MG_State::GLState::ProgramObject& program,
                                             const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
-                                            VkDescriptorBufferInfo& outBufferInfo) const;
+                                            Uint32 element, VkDescriptorBufferInfo& outBufferInfo) const;
+        // `element` indexes an image ARRAY inside one binding; each element carries its own
+        // independently assigned GL image unit.
         Bool ResolveStorageImageDescriptor(VkCommandBuffer commandBuffer,
                                            const MG_State::GLState::ProgramObject& program,
                                            const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
-                                           VkDescriptorImageInfo& outImageInfo) const;
+                                           Uint32 element, VkDescriptorImageInfo& outImageInfo) const;
         // Result of resolving a UBO binding: either a zero-copy direct bind to the app's resident
         // VkBuffer (the GLES backend's approach - no per-draw copy) or the CPU payload to upload.
         struct UboBindResult {
@@ -149,8 +227,23 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Bool ResolveUniformBufferPayload(const MG_State::GLState::ProgramObject& program,
                                          const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
                                          Uint32 arrayElement, UboBindResult& out) const;
-        Bool CreateDescriptorPool(Uint32 maxSets, VkDescriptorPool& outPool) const;
+        // Shared resolution of one dynamic-UBO binding element into the
-        Bool GrowFrameDescriptorPool(FrameResources& frame, Uint32 frameIndex);
+        // (buffer, range, dynamicOffset) triple the descriptor consumes: direct
+        // bind, global-slice reuse, or transient upload. Used by the full walk
+        // and by the dynamic-offset-only rebind (see FastRebindMemo).
+        Bool ResolveDynamicUboDescriptor(const MG_State::GLState::ProgramObject& program,
+                                         const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
+                                         Uint32 arrayElement, Uint32 frameIndex, VkBuffer& outBuffer,
+                                         VkDeviceSize& outRange, Uint32& outDynamicOffset);
+        // The vkCmdBindDescriptorSets tail shared by the full walk and the
+        // dynamic-offset-only rebind: skips the driver call when this exact
+        // binding is already live on the command buffer (see the bind-dedup
+        // shadow below), otherwise binds and refreshes the shadow.
+        void BindDescriptorSetDeduped(VkCommandBuffer commandBuffer, VkPipelineBindPoint bindPoint,
+                                      VkPipelineLayout pipelineLayout, VkDescriptorSet descriptorSet,
+                                      const Vector<Uint32>& dynamicOffsets);
+        Bool CreateDescriptorPool(Uint32 maxSets, Bool updateAfterBind, VkDescriptorPool& outPool) const;
+        Bool GrowFrameDescriptorPool(FrameResources& frame, Uint32 frameIndex, Bool updateAfterBind);
         VkResult AllocateDescriptorSetsFromActivePool(
             Uint32 frameIndex, const ProgramFactory::VkProgramObject& programObj, VkDescriptorSet& outDescriptorSet);
         VkResult AcquireDescriptorSet(Uint32 frameIndex,
@@ -179,14 +272,53 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Vector<VkBufferView> m_texelBufferViewsScratch;
         Vector<Uint32> m_dynamicOffsetsScratch;
 
-        // Descriptor-set reuse across consecutive draws (see BindProgramUniformBuffers).
+        // Descriptor-set reuse across recent draws (see BindProgramUniformBuffers).
-        // When a draw's resolved descriptor content is byte-identical to the previous
+        // When a draw's resolved descriptor content is byte-identical to one memoized
-        // draw's, reuse the same VkDescriptorSet and skip AcquireDescriptorSet +
+        // earlier, reuse that VkDescriptorSet and skip AcquireDescriptorSet +
-        // vkUpdateDescriptorSets - only the bind-time dynamic offsets differ. Reset each
+        // vkUpdateDescriptorSets - only the bind-time dynamic offsets differ. Four
-        // frame in BeginFrame because the frame's descriptor sets are recycled there.
+        // entries with round-robin replacement rather than one: draws alternating
-        VkDescriptorSet m_lastBoundDescriptorSet = VK_NULL_HANDLE;
+        // between two programs (MC's chunk<->entity ping-pong) would thrash a single
-        Uint64 m_lastDescriptorSignature = 0;
+        // slot into a full re-allocate+write every draw. Reset each frame in BeginFrame
-        Bool m_hasLastDescriptor = false;
+        // because the frame's descriptor sets are recycled there.
+        struct DescriptorReuseEntry {
+            Uint64 signature = 0;
+            VkDescriptorSet set = VK_NULL_HANDLE;
+            Bool valid = false;
+        };
+        static constexpr Uint32 kDescriptorReuseMemoSize = 4;
+        DescriptorReuseEntry m_descriptorReuseMemo[kDescriptorReuseMemoSize];
+        Uint32 m_descriptorReuseMemoNext = 0;
+
+        // Dynamic-offset-only rebind (see BindProgramUniformBuffers): records the
+        // descriptor set selected by the last cacheable full walk of a program
+        // whose active bindings are exactly one dynamic UBO (single descriptor)
+        // plus combined-image samplers. When the next call proves every sampler
+        // descriptor input unchanged (samplerDescriptorsUnchangedHint) and the
+        // UBO re-resolves to the SAME VkBuffer+range - only the dynamic offset
+        // moved, the per-draw glUniform case - the walk collapses to: resolve one
+        // offset, rebind the recorded set with new pDynamicOffsets (Vulkan allows
+        // rebinding the same set with different dynamic offsets).
+        // Invalidation inventory: BeginFrame clears it (the frame's sets are
+        // recycled) and the frameIndex field guards cross-frame confusion on top;
+        // OnDescriptorSetLayoutDestroyed clears it (the set may be freed); a
+        // sampler-override walk clears it (mirrors m_descriptorReuseMemo); a
+        // program relink bumps the backend state version and thus programObj.hash
+        // so the key misses; the program lifetime id is never reused, so a
+        // deleted-and-recreated program misses; a texture/sampler/binding change
+        // drops the hint upstream; an arena wrap or growth resolves a different
+        // VkBuffer and misses. AcquireDescriptorSet's per-frame cursor only
+        // advances, so the recorded set is never re-written within its frame.
+        struct FastRebindMemo {
+            Bool valid = false;
+            Uint32 frameIndex = 0;
+            Uint64 programLifetimeId = 0;
+            ProgramFactory::HashType programHash = 0;
+            Uint32 uboBinding = 0;
+            VkBuffer uboBuffer = VK_NULL_HANDLE;
+            VkDeviceSize uboRange = 0;
+            VkDescriptorSet set = VK_NULL_HANDLE;
+        };
+        FastRebindMemo m_fastRebindMemo;
 
         // vkCmdBindDescriptorSets dedup: consecutive draws with a static uniform
         // block resolve to the same set AND the same dynamic offsets, so the
@@ -228,8 +360,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // lifetime id, so a freed-and-reallocated sampler or texture at the same heap address
         // always gets a fresh id and misses (a raw pointer would false-hit that ABA) - so a
         // stale guess can only miss and fall through to the hash, never resolve wrong. Still
-        // reset each frame alongside the descriptor-set cache. Indexed by binding.
+        // reset each frame alongside the descriptor-set cache. Indexed by binding, but the
+        // whole-descriptor entry is additionally keyed by program lifetime: Vulkan binding
+        // numbers are layout-local and unrelated programs routinely reuse binding 0/1.
         struct SamplerResolveMemo {
+            Uint64 infoProgramLifetimeId = 0;
             Uint64 samplerLifetimeId = 0;
             Uint64 textureLifetimeId = 0;
             VkSampler sampler = VK_NULL_HANDLE;
@@ -245,7 +380,36 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             SamplerNumericDomain viewFormatDomain = SamplerNumericDomain::Unknown;
             VkFormat viewFormat = VK_FORMAT_UNDEFINED;
             Bool viewFormatValid = false;
+            // Whole resolved descriptor from this binding's last full resolve. Reused
+            // ONLY under ResolveSamplerDescriptor's trustUnchangedHint, whose caller
+            // proves every resolve input unchanged; cleared with the per-frame reset
+            // (the cached VkSampler outlives a frame only via a fresh resolve, which
+            // also re-stamps it against VkSamplerManager's frame-boundary sweep).
+            //
+            // This one field is keyed by binding but describes ONE descriptor, so it is
+            // written and read only for single-descriptor bindings. A sampler ARRAY's
+            // elements share the binding and would overwrite each other here - the last
+            // element resolved would then be handed to element 0 on the next hinted draw.
+            // Every other field above is self-validating (each compares its full key
+            // before reuse, and the view-format entry is a pure function of format and
+            // numeric domain), so an arrayed binding may keep using those.
+            VkDescriptorImageInfo info{};
+            Bool infoValid = false;
         };
         mutable Vector<SamplerResolveMemo> m_samplerResolveMemo;
+        // Exclusive upper bound on the entries of m_samplerResolveMemo that any resolve
+        // has ever written. The vector is sized to the DEVICE binding cap (256 on desktop
+        // NVIDIA), but a program declares 1-8 bindings, so the per-frame reset below was
+        // memsetting ~22 KB of never-touched entries every frame - a measurable slice of
+        // the per-frame fixed cost on draw-light frames. Every site that can turn any of
+        // an entry's *Valid flags on raises this mark first, so entries at or above it are
+        // provably still in their constructed (all-invalid) state and clearing them is a
+        // no-op. Never lowered except by Initialize/Shutdown, which rebuild the vector.
+        mutable Uint32 m_samplerResolveMemoHighWater = 0;
+        void NoteSamplerResolveMemoTouched(Uint32 binding) const {
+            if (binding >= m_samplerResolveMemoHighWater) {
+                m_samplerResolveMemoHighWater = binding + 1;
+            }
+        }
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VertexInputStateFactory.cpp

@@ -29,17 +29,22 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             XXHASH_VERIFY(XXH64_update(m_hashState, &attr.Stride, sizeof(attr.Stride)));
             XXHASH_VERIFY(XXH64_update(m_hashState, &attr.Offset, sizeof(attr.Offset)));
             XXHASH_VERIFY(XXH64_update(m_hashState, &attr.IsInteger, sizeof(attr.IsInteger)));
+            XXHASH_VERIFY(XXH64_update(m_hashState, &attr.IsLong, sizeof(attr.IsLong)));
             XXHASH_VERIFY(XXH64_update(m_hashState, &attr.IsBgra, sizeof(attr.IsBgra)));
             XXHASH_VERIFY(XXH64_update(m_hashState, &attr.Divisor, sizeof(attr.Divisor)));
 
-            // The buffer's heap address is an identity component of the key: a freed
+            // The bound buffer's IDENTITY is a component of the key, and it has to be the
-            // buffer's reused address can alias an old cache entry, but only under a
+            // buffer's never-reused lifetime id - NOT its heap address, which this used to
-            // byte-identical attribute layout - and the entry payload is a pure function
+            // hash. An address is recycled by the allocator, so a deleted-and-recreated
-            // of the hashed inputs, with the draw path re-resolving bindingBufferKeys
+            // buffer reproduces it; combined with a byte-identical attribute layout that
-            // against the live VAO attribute pointers, so an aliased hit returns exactly
+            // reproduces the WHOLE content hash, and the hash is what
-            // what a rebuild would. Address drift only grows the map; the OnFrameBoundary
+            // TryBindResolvedVertexBindings accepts as proof that a memoised binding still
-            // aging sweep bounds that.
+            // reads the buffer it was resolved from. It did not: a destroyed buffer's GPU
-            const SizeT bufferKey = reinterpret_cast<SizeT>(attr.Buffer.get());
+            // slice was bound for its successor's draw, which is how a transform-feedback
+            // capture came back holding a dead VAO's vertex data (0,0,0,1 - the previous
+            // test's positions) instead of its own.
+            // Zero for client memory (no buffer), which is a distinct identity of its own.
+            const Uint64 bufferKey = attr.Buffer ? attr.Buffer->GetLifetimeId() : 0;
             XXHASH_VERIFY(XXH64_update(m_hashState, &bufferKey, sizeof(bufferKey)));
         }
 
@@ -70,6 +75,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
         const BackendVertexInputState& entry = GetOrCreateVertexInputState(vao, GetOrComputeHash(vao));
         vao.SetBackendStateMemo(&entry, m_evictionEpoch);
+        // Also mirror the layout identity and the two per-draw masks into the VAO's aux
+        // memo (pure VALUES derived from the VAO configuration, so config-version
+        // guarding alone is sound). The draw fast path reads them from the VAO object it
+        // already touched instead of chasing into this entry - see PackVertexInputAuxMemo.
+        vao.SetBackendAuxMemo(entry.layoutHash,
+                              PackVertexInputAuxMasks(entry.unsupportedAttribMask, entry.attributeLocationMask));
         return entry;
     }
 
@@ -97,9 +108,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             }
 
             const VkFormat sourceVkFormat =
-                ToVkVertexFormat(attr.Type, attr.Size, attr.Normalized, attr.IsInteger, attr.IsBgra);
+                ToVkVertexFormat(attr.Type, attr.Size, attr.Normalized, attr.IsInteger, attr.IsBgra, attr.IsLong);
             if (sourceVkFormat == VK_FORMAT_UNDEFINED) {
-                MGLOG_E("Unsupported vertex attribute layout (location=%u, type=%s, size=%d): the array is "
+                MGLOG_E_ONCE("Unsupported vertex attribute layout (location=%u, type=%s, size=%d): the array is "
                         "enabled but cannot be mapped to a VkFormat",
                         location, MG_Util::ConvertDataTypeToString(attr.Type).c_str(), attr.Size);
                 unsupportedAttribMask |= (1u << location);
@@ -114,7 +125,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                     if (fallbackFormat != VK_FORMAT_UNDEFINED && SupportsVertexBufferFormat(fallbackFormat)) {
                         vkFormat = fallbackFormat;
                         conversion = VertexStreamConversion::ScaledIntegerToFloat32;
-                        MGLOG_W("Vertex attribute location=%u format=%d lacks "
+                        MGLOG_W_ONCE("Vertex attribute location=%u format=%d lacks "
                                 "VK_FORMAT_FEATURE_VERTEX_BUFFER_BIT; using float32 stream format=%d "
                                 "(type=%s size=%d normalized=%s integer=%s)",
                                 location, static_cast<Int>(sourceVkFormat), static_cast<Int>(vkFormat),
@@ -124,7 +135,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 }
 
                 if (conversion == VertexStreamConversion::None) {
-                    MGLOG_E("Unsupported Vulkan vertex format (location=%u, format=%d, type=%s, size=%d): "
+                    MGLOG_E_ONCE("Unsupported Vulkan vertex format (location=%u, format=%d, type=%s, size=%d): "
                             "VK_FORMAT_FEATURE_VERTEX_BUFFER_BIT is unavailable and no semantic fallback exists",
                             location, static_cast<Int>(sourceVkFormat),
                             MG_Util::ConvertDataTypeToString(attr.Type).c_str(), attr.Size);
@@ -135,15 +146,21 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
             const SizeT attribByteSize = GetAttributeByteSize(attr.Type, attr.Size, attr.IsBgra);
             if (attribByteSize == 0) {
-                MGLOG_E("Vertex attribute with unknown component size (location=%u, type=%s): the array is "
+                MGLOG_E_ONCE("Vertex attribute with unknown component size (location=%u, type=%s): the array is "
                         "enabled but cannot be sized",
                         location, MG_Util::ConvertDataTypeToString(attr.Type).c_str());
                 unsupportedAttribMask |= (1u << location);
                 continue;
             }
 
-            const Uint32 sourceStride =
+            // Verbatim, zero included. The frontend already resolved a pointer call's
-                attr.Stride > 0 ? static_cast<Uint32>(attr.Stride) : static_cast<Uint32>(attribByteSize);
+            // "tightly packed" stride 0 into the element size (see VertexAttribute::Stride),
+            // so a zero here is the binding model's stride 0 - every vertex reads the same
+            // element - which is exactly what a zero VkVertexInputBindingDescription::stride
+            // means. Substituting the element size fetched a fresh element per vertex and ran
+            // off the end of the buffer (KHR-GL43.vertex_attrib_binding.basic-input-case7/8).
+            // Client-memory arrays cannot reach zero: they only exist on the pointer path.
+            const Uint32 sourceStride = static_cast<Uint32>(attr.Stride);
             const Bool packedAttribute = attr.Type == DataType::Int2101010Rev ||
                                          attr.Type == DataType::Uint2101010Rev;
             const SizeT requiredAlignment = packedAttribute ? attribByteSize : GetComponentSize(attr.Type);
@@ -158,17 +175,23 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 // unless VK_EXT_legacy_vertex_attributes is available, so deinterleave this one
                 // attribute into a tightly packed transient stream without changing its format.
                 conversion = VertexStreamConversion::Repack;
-                MGLOG_W("Vertex attribute location=%u uses Vulkan-incompatible alignment "
+                MGLOG_W_ONCE("Vertex attribute location=%u uses Vulkan-incompatible alignment "
                         "(offset=%zu stride=%u required=%zu); using a tightly packed stream",
                         location, attr.Offset, sourceStride, requiredAlignment);
             }
 
             Uint32 stride = sourceStride;
+            // A converted stream is tightly packed, so its stride is the converted element
+            // size - unless the source stride is zero, which does not describe a packing at
+            // all but "never advance". That survives the conversion unchanged: the draw path
+            // converts exactly one element and every vertex reads it.
+            if (sourceStride != 0) {
                 if (conversion == VertexStreamConversion::Repack) {
                     stride = static_cast<Uint32>(attribByteSize);
                 } else if (conversion == VertexStreamConversion::ScaledIntegerToFloat32) {
                     stride = static_cast<Uint32>(attr.Size * static_cast<Int>(sizeof(Float)));
                 }
+            }
             const VkVertexInputRate inputRate =
                 (attr.Divisor == 0) ? VK_VERTEX_INPUT_RATE_VERTEX : VK_VERTEX_INPUT_RATE_INSTANCE;
 
@@ -264,8 +287,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             if (m_frameBoundaryCounter - it->second->lastUsedFrameBoundary > kRetireAgeBoundaries) {
                 it = m_cache.erase(it);
                 // Invalidate every VAO's state-pointer memo: the erased node's
-                // address may be reused by a future insert.
+                // address may be reused by a future insert. Advance through the
-                ++m_evictionEpoch;
+                // process-wide source so the value stays unique across factory
+                // instances (see the member comment).
+                m_evictionEpoch = ++s_evictionEpochSource;
             } else {
                 ++it;
             }
@@ -273,7 +298,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     }
 
     VkFormat VertexInputStateFactory::ToVkVertexFormat(DataType type, Int size, Bool normalized, Bool isInteger,
-                                                       Bool isBgra) {
+                                                       Bool isBgra, Bool isLong) {
         if (isBgra) {
             // GL_BGRA: four reversed-order components, always normalized (enforced at validation), only
             // legal with GL_UNSIGNED_BYTE or a 2_10_10_10 type. The reversed VkFormats put the
@@ -298,6 +323,22 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         case DataType::Int2101010Rev:
             if (isInteger || size != 4) return VK_FORMAT_UNDEFINED;
             return normalized ? VK_FORMAT_A2B10G10R10_SNORM_PACK32 : VK_FORMAT_A2B10G10R10_SSCALED_PACK32;
+        case DataType::Float64:
+            // A 64-bit attribute is fetched as its 32-bit word pair and bitcast back to double in the
+            // shader (PackDoubleVertexInputsPass does the shader half). That is bit-exact and, unlike
+            // VK_FORMAT_R64*_SFLOAT, needs no format capability: lavapipe reports bufferFeatures = 0
+            // for every R64 float format, so a native 64-bit vertex fetch is simply unavailable there
+            // while shaderFloat64 is not. Both halves key off nothing but the attribute being long,
+            // so they always agree without extra plumbing.
+            if (!isLong || isInteger || normalized) return VK_FORMAT_UNDEFINED;
+            switch (size) {
+            case 1: return VK_FORMAT_R32G32_UINT;
+            case 2: return VK_FORMAT_R32G32B32A32_UINT;
+            // A dvec3/dvec4 input is 6/8 uint32 components: no single VkFormat, and GL spreads it
+            // over two attribute locations, which the location-per-VAO-index model here does not
+            // express. Declined rather than fetched wrong.
+            default: return VK_FORMAT_UNDEFINED;
+            }
         case DataType::Float32:
             switch (size) {
             case 1: return VK_FORMAT_R32_SFLOAT;

MobileGL/MG_Backend/DirectVulkan/Renderer/VertexInputStateFactory.h

@@ -28,11 +28,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         struct BackendVertexInputState {
             HashType hash = 0;
             // Hash of the resolved Vulkan vertex layout only (bindings, attributes,
-            // unsupported mask) - NO buffer identities. `hash` mixes buffer heap
+            // unsupported mask) - NO buffer identities. `hash` mixes each bound
-            // addresses so per-chunk VBOs mint a fresh identity per buffer; keying
+            // buffer's never-reused LIFETIME ID, so per-chunk VBOs mint a fresh
-            // pipelines on that minted one VkPipeline per chunk section for an
+            // identity per buffer; keying pipelines on that minted one VkPipeline per
-            // identical layout, defeating pipeline reuse and the per-draw memo.
+            // chunk section for an identical layout, defeating pipeline reuse and the
-            // Pipelines depend only on the layout, so they key on this instead.
+            // per-draw memo. Pipelines depend only on the layout, so they key on this
+            // instead.
             HashType layoutHash = 0;
             // Frame boundary of the last cache hit; entries idle past the
             // OnFrameBoundary retirement age are evicted (CPU heap only).
@@ -71,6 +72,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         ~VertexInputStateFactory() = default;
         VertexInputStateFactory(const VertexInputStateFactory&) = delete;
 
+        // The VAO aux-memo payload GetOrCreateVertexInputState(vao) stamps: aux0 is the
+        // entry's layoutHash, aux1 packs (unsupportedAttribMask << 32) | attributeLocationMask.
+        // Readers that find the aux memo valid can use these without resolving the entry.
+        static Uint64 PackVertexInputAuxMasks(Uint32 unsupportedAttribMask, Uint32 attributeLocationMask) {
+            return (static_cast<Uint64>(unsupportedAttribMask) << 32) | attributeLocationMask;
+        }
+
         HashType ComputeHash(const MG_State::GLState::VertexArrayObject& vao) const;
         // Memoized ComputeHash: reuses the VAO's cached hash while its config version
         // is unchanged. Use this on per-draw paths.
@@ -79,8 +87,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             const MG_State::GLState::VertexArrayObject& vao, HashType hash);
         const BackendVertexInputState& GetOrCreateVertexInputState(const MG_State::GLState::VertexArrayObject& vao);
         // Frame boundary hook: ages the cache and evicts entries not hit for many
-        // frames. The key mixes buffer heap addresses, so buffer/VAO churn keeps
+        // frames. The key mixes each bound buffer's never-reused lifetime id, so
-        // minting fresh keys; without eviction the map grows for the whole session.
+        // buffer/VAO churn keeps minting fresh keys - and does so by construction,
+        // not by luck: a recreated buffer can no longer land back on its dead
+        // predecessor's key. Without eviction the map grows for the whole session.
         // Entries hold no Vulkan handles (pipeline creation copies the descriptions)
         // and the draw path's entry reference never spans a frame boundary, so
         // eviction here needs no GPU-idle proof. Self-gated: one counter bump and
@@ -93,17 +103,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         static SizeT GetAttributeByteSize(DataType type, Int size, Bool isBgra);
 
     private:
-        static VkFormat ToVkVertexFormat(DataType type, Int size, Bool normalized, Bool isInteger, Bool isBgra = false);
+        static VkFormat ToVkVertexFormat(DataType type, Int size, Bool normalized, Bool isInteger, Bool isBgra = false,
+                                         Bool isLong = false);
         static Bool IsScaledIntegerVertexFormat(VkFormat format);
         static VkFormat ToFloat32VertexFormat(Int componentCount);
         Bool SupportsVertexBufferFormat(VkFormat format) const;
 
         const VulkanRendererConfig& m_config;
         VkPhysicalDevice m_physicalDevice = VK_NULL_HANDLE;
-        // Values are heap-allocated: FastSTL::unordered_map is open-addressing,
+        // Values are heap-allocated: UnorderedMap is open-addressing, so INSERT
-        // so INSERT invalidates references to stored values. The draw path (and
+        // invalidates references to stored values - and so does ERASE, which shifts
-        // the VAOs' state-pointer memos) hold entry pointers across inserts;
+        // the rest of the probe cluster into the hole and therefore moves entries
-        // only the unique_ptr cell moves, never the pointee.
+        // other than the erased one. The draw path (and the VAOs' state-pointer
+        // memos) hold entry pointers across both; only the unique_ptr cell moves,
+        // never the pointee.
         UnorderedMap<HashType, UniquePtr<BackendVertexInputState>> m_cache;
         // Monotonic frame-boundary counter (bumped in OnFrameBoundary) for cache aging.
         Uint64 m_frameBoundaryCounter = 0;
@@ -112,7 +125,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // construction); a memo is honored only while its recorded epoch
         // matches, so an evicted entry can never be dereferenced through a
         // stale memo.
-        Uint64 m_evictionEpoch = 1;
+        //
+        // Drawn from a process-wide source, never a per-instance counter: the VAO
+        // memos outlive this factory (they live on pGLContext's VAOs, the renderer
+        // is destroyed and recreated on EGL surface release/re-create), so a fresh
+        // factory restarting at a dead factory's epoch value would honor its
+        // dangling entry pointers. The constructor takes a value strictly greater
+        // than anything a predecessor ever stamped, so a dead factory's memo can
+        // never compare equal here - the same never-reused idiom as the lifetime ids.
+        // Single-threaded like the rest of the factory (renderer-thread only).
+        static inline Uint64 s_evictionEpochSource = 0;
+        Uint64 m_evictionEpoch = ++s_evictionEpochSource;
         static inline XXH64_state_t* m_hashState = XXH64_createState();
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferManager.cpp

@@ -23,7 +23,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT |
             VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
             VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT | VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT |
-            VK_BUFFER_USAGE_TRANSFER_SRC_BIT;
+            // "Every usage" has to mean every usage: a buffer texture reached through an IMAGE
+            // unit takes a VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER descriptor, and the write is
+            // invalid unless the buffer was created with this bit. Nothing asked for it until
+            // imageBuffer support existed, so the omission was invisible.
+            VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_SRC_BIT;
         // Appended to kPersistentBackedUsage when VK_EXT_transform_feedback is enabled
         // (see VkBufferManagerInitInfo::transformFeedbackUsageEnabled).
         constexpr VkBufferUsageFlags kTransformFeedbackUsage =
@@ -161,12 +165,23 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     }
 
     void VkBufferManager::CollectAllDeferredReleases() {
+        // Per-resource releases only. Every one of them was deferred behind a BumpSliceEpoch,
+        // so no memo can still name the handle, and the caller has proved the GPU is idle.
+        //
+        // The transient arena's releases are deliberately NOT collected here. A buffer lands
+        // there when the arena outgrows it mid-frame (BufferArena::EnsureCapacity), and at
+        // that moment every slice already handed out from this frame's arena still names it -
+        // VkBufferResource::transientSlice above all, which AcquireStreamedSlice keeps
+        // serving for the whole frame serial on the strength of transientFrameSerial alone.
+        // Nothing bumps the slice epoch for those other resources, so freeing the buffer
+        // here left the streamed memo handing a destroyed VkBuffer to vkCmdBindIndexBuffer
+        // (llvmpipe then faulted inside the draw; the Create/Flywheel indirect retrace died
+        // exactly this way). Mid-frame drains do not advance m_frameSerial, so they must not
+        // free arena storage either: the arena's own ResetFrame/BeginFrame is the point where
+        // the slot's slices stop being reachable, and that is where these releases land.
         for (Uint32 frameIndex = 0; frameIndex < m_deferredBufferReleases.size(); ++frameIndex) {
             CollectDeferredReleases(frameIndex);
         }
-        for (Uint32 frameIndex = 0; frameIndex < m_transientUploadArena.GetFrameCount(); ++frameIndex) {
-            m_transientUploadArena.CollectDeferredReleases(frameIndex);
-        }
     }
 
     void VkBufferManager::NotifyDeviceIdle() {
@@ -242,8 +257,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     }
 
     void VkBufferManager::TrackLiveResource(const SharedPtr<VkBufferResource>& resource) {
-        if (m_liveResources.size() >= kLiveResourcePruneThreshold) {
+        // Sweep on a doubling watermark rather than on every insert past the threshold. The old
+        // form walked the whole vector for each new buffer once the list passed 256, and when the
+        // buffers are all live the walk removes nothing and the list grows by one - so creating N
+        // live buffers cost ~N^2/2 expired() checks. Reclamation semantics are unchanged: the sweep
+        // still removes exactly the expired entries, just less often and with the same bound on how
+        // much dead weight can accumulate (at most as many entries as were live at the last sweep).
+        if (m_liveResources.size() >= std::max<SizeT>(kLiveResourcePruneThreshold, 2 * m_liveResourcesLastPruned)) {
             std::erase_if(m_liveResources, [](const WeakPtr<VkBufferResource>& weak) { return weak.expired(); });
+            m_liveResourcesLastPruned = m_liveResources.size();
         }
         m_liveResources.push_back(resource);
     }
@@ -251,6 +273,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     void VkBufferManager::ReleaseAllLiveResources() {
         for (auto& weak : m_liveResources) {
             if (auto resource = weak.lock()) {
+                BumpSliceEpoch(*resource);
                 resource->buffer.Destroy();
                 resource->storageSize = 0;
                 resource->usageFlags = 0;
@@ -265,6 +288,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
     Bool VkBufferManager::CreateResidentStorage(VkBufferResource& resource, VkDeviceSize size,
                                                 VkBufferUsageFlags usage, VkMemoryPropertyFlags requiredFlags) {
+        // The only place a resident VkBuffer handle is minted, so every resident slice
+        // change funnels through here (callers release the old handle first).
+        BumpSliceEpoch(resource);
         // Staged range copies write resident storage with vkCmdCopyBuffer.
         usage |= VK_BUFFER_USAGE_TRANSFER_DST_BIT;
         const Bool created = resource.buffer.Create({
@@ -276,7 +302,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             .requiredFlags = requiredFlags,
         });
         if (!created || resource.buffer.Map() == nullptr) {
-            MGLOG_E("VkBufferManager::CreateResidentStorage failed (size=%llu)",
+            MGLOG_E_ONCE("VkBufferManager::CreateResidentStorage failed (size=%llu)",
                     static_cast<unsigned long long>(size));
             resource.buffer.Destroy();
             resource.storageSize = 0;
@@ -298,7 +324,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return false;
         }
         if (!resource.buffer.Upload(bufferObject.MappedData(), size, 0)) {
-            MGLOG_E("VkBufferManager::SwapStorageAndUploadAll: upload failed");
+            MGLOG_E_ONCE("VkBufferManager::SwapStorageAndUploadAll: upload failed");
             resource.pendingFullUpload = true;
             return false;
         }
@@ -351,8 +377,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (!resource) {
             return; // lazy: AcquireResidentSlice performs a full upload on creation
         }
+        // A respecify can change the size, the usage hint (so the resident/streamed
+        // route), and the contents at once; retire every memo before deciding what to
+        // do about the storage.
+        BumpSliceEpoch(*resource);
         // Any cached streaming slice refers to the previous contents.
         resource->transientFrameSerial = 0;
+        // Redefining the store hands any adopted mapping back to the CPU shadow
+        // (BufferObject::RedefineStorage), so a buffer that reaches here persistent-mapped
+        // is an ordinary resident one again: it needs the busy-tracking and conditional
+        // orphan below, and the next AcquirePersistentMap has to mint storage for the new
+        // store rather than hand back a mapping of the old one.
+        resource->persistentMapped = false;
         if (!resource->buffer.IsValid()) {
             return; // streaming-only resource: shadow + serial are enough
         }
@@ -373,7 +409,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
 
         if (!resource->buffer.Upload(bufferObject.MappedData(), size, 0)) {
-            MGLOG_E("VkBufferManager::OnRespecify: in-place upload failed");
+            MGLOG_E_ONCE("VkBufferManager::OnRespecify: in-place upload failed");
             resource->pendingFullUpload = true;
         }
     }
@@ -383,6 +419,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (!resource) {
             return;
         }
+        // Drops the streaming memo below and may end in a storage swap or a deferred
+        // full re-upload, so no memoised slice survives this.
+        BumpSliceEpoch(*resource);
         resource->transientFrameSerial = 0;
         if (!resource->buffer.IsValid() || resource->pendingFullUpload) {
             return;
@@ -395,7 +434,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (!IsResourceBusy(*resource)) {
             if (!resource->buffer.Upload(bufferObject.MappedData() + offset,
                                          static_cast<VkDeviceSize>(size), static_cast<VkDeviceSize>(offset))) {
-                MGLOG_E("VkBufferManager::OnSubData: host upload failed");
+                MGLOG_E_ONCE("VkBufferManager::OnSubData: host upload failed");
                 resource->pendingFullUpload = true;
             }
             return;
@@ -415,6 +454,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (!resource) {
             return;
         }
+        BumpSliceEpoch(*resource);
         resource->transientFrameSerial = 0;
         if (!resource->buffer.IsValid() || resource->pendingFullUpload) {
             return;
@@ -431,7 +471,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if ((appAccess & BufferMappingAccessBit::Unsynchronized) || !IsResourceBusy(*resource)) {
             if (!resource->buffer.Upload(bufferObject.MappedData() + offset,
                                          static_cast<VkDeviceSize>(size), static_cast<VkDeviceSize>(offset))) {
-                MGLOG_E("VkBufferManager::OnFlushMappedRange: host upload failed");
+                MGLOG_E_ONCE("VkBufferManager::OnFlushMappedRange: host upload failed");
                 resource->pendingFullUpload = true;
             }
             return;
@@ -474,6 +514,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             TrackLiveResource(resource);
         }
 
+        // Bumped for the request, not just for the storage it may create. This is the
+        // one call the frontend makes when a buffer becomes persistently mapped for
+        // writing (BufferObject::AcquireMemoryRange), and a map the backend declines
+        // keeps mutating its shadow with no further API call - so it is what lets
+        // GetSliceEpochCounter stand for "no buffer needs a persistent-map range push".
+        BumpSliceEpoch(*resource);
+
         // Idempotent: an already-backed buffer returns the same mapped base.
         if (resource->persistentMapped && resource->buffer.IsValid() && resource->storageSize == size) {
             return resource->buffer.GetMappedData();
@@ -516,7 +563,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         const VkDeviceSize size = static_cast<VkDeviceSize>(bufferObject->GetSize());
         if (size == 0) {
-            MGLOG_E("VkBufferManager::AcquireResidentSlice failed: buffer size is zero");
+            MGLOG_E_ONCE("VkBufferManager::AcquireResidentSlice failed: buffer size is zero");
             return false;
         }
 
@@ -538,7 +585,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 return false;
             }
             if (!resource->buffer.Upload(bufferObject->MappedData(), size, 0)) {
-                MGLOG_E("VkBufferManager::AcquireResidentSlice failed: initial upload failed");
+                MGLOG_E_ONCE("VkBufferManager::AcquireResidentSlice failed: initial upload failed");
                 resource->buffer.Destroy();
                 resource->storageSize = 0;
                 resource->usageFlags = 0;
@@ -573,7 +620,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         const VkDeviceSize size = static_cast<VkDeviceSize>(bufferObject->GetSize());
         if (size == 0) {
-            MGLOG_E("VkBufferManager::AcquireStreamedSlice failed: buffer size is zero");
+            MGLOG_E_ONCE("VkBufferManager::AcquireStreamedSlice failed: buffer size is zero");
             return false;
         }
 
@@ -584,6 +631,41 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return true;
         }
 
+        // Idle-content promotion: see the field comments in VkBufferResource. The
+        // streak counts frame BOUNDARIES survived unchanged (the same-frame memo
+        // above swallows repeat draws), so a promotion needs the content stable
+        // for kStreamedPromotionStreak whole frames - one no-op frame does not
+        // trigger the resident round-trip, whose creation upload is itself a
+        // staged copy worth avoiding for content that is about to change again.
+        constexpr Uint32 kStreamedPromotionStreak = 2;
+        if (resource->promotedResident) {
+            if (resource->promotedChangeSerial == changeSerial &&
+                static_cast<VkDeviceSize>(bufferObject->GetSize()) == size) {
+                return AcquireResidentSlice(kind, bufferObject, outSlice);
+            }
+            resource->promotedResident = false;
+            resource->unchangedStreak = 0;
+        } else if (resource->transientChangeSerial == changeSerial && resource->transientSize == size &&
+                   resource->transientFrameSerial != 0) {
+            if (++resource->unchangedStreak >= kStreamedPromotionStreak) {
+                // Promotion moves the buffer off the arena and onto resident storage.
+                resource->promotedResident = true;
+                resource->promotedChangeSerial = changeSerial;
+                BumpSliceEpoch(*resource);
+                if (AcquireResidentSlice(kind, bufferObject, outSlice)) {
+                    return true;
+                }
+                resource->promotedResident = false; // resident creation failed: stream as before
+            }
+        } else {
+            resource->unchangedStreak = 0;
+        }
+
+        // A fresh arena allocation: a different slice than the last call handed back,
+        // and (below) the point where a promoted buffer's resident storage is released.
+        // The stable-promotion exit above returns before this, so a buffer the app has
+        // stopped touching keeps one slice for as long as it keeps its resident storage.
+        BumpSliceEpoch(*resource);
         if (!m_transientUploadArena.Upload(m_currentFrameIndex, bufferObject->MappedData(), size, 16,
                                            outSlice)) {
             return false;
@@ -636,7 +718,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         case BufferKind::Uniform:
             return VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT;
         case BufferKind::TextureBuffer:
-            return VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT;
+            // Both texel roles, for the same reason vertex/index carry both bits: one GL buffer
+            // texture can be read as a samplerBuffer and written as an imageBuffer, and which of
+            // the two it is only becomes known when a shader that uses it is bound - long after
+            // the resident buffer was created. A VkBufferView for a storage-texel descriptor is
+            // invalid unless the buffer was created with the storage bit, so a buffer that
+            // acquired only the uniform bit could never be given one.
+            return VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT;
         case BufferKind::ShaderStorage:
             return VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT;
         case BufferKind::Indirect:

MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferManager.h

@@ -57,11 +57,33 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // never orphaned or recreated. Draw-time acquire binds it directly, no re-upload.
         Bool persistentMapped = false;
 
+        // Bumped from a manager-wide counter every time anything that decides which
+        // BufferSlice an Acquire*Slice call hands back changes: storage created or
+        // released, a full re-upload becoming due, a promotion/demotion between
+        // resident and streamed storage, or a new per-frame arena slice. Callers that
+        // memoise a resolved slice compare this to prove the memo still describes the
+        // buffer. The counter is manager-wide (never per-resource) so a freshly
+        // created resource - including one that replaces a destroyed resource at the
+        // same address - can never reproduce a value some memo already holds. 0 means
+        // "no slice has ever been handed out", which no memo can match.
+        Uint64 sliceEpoch = 0;
+
         // Cached transient (streaming) slice for the current frame.
         BufferSlice transientSlice{};
         Uint64 transientFrameSerial = 0;
         Uint64 transientChangeSerial = 0;
         VkDeviceSize transientSize = 0;
+
+        // Streaming re-copies the whole store into the per-frame arena on every
+        // frame, which is right for genuinely per-frame data but pure waste for a
+        // Dynamic-hinted buffer the app stopped touching. After the content
+        // survives kStreamedPromotionStreak frame boundaries unchanged it is
+        // promoted to resident storage (one final upload, then zero per-frame
+        // cost); the first content change demotes it back to streaming, and the
+        // streaming path's existing downgrade releases the resident store.
+        Uint32 unchangedStreak = 0;
+        Bool promotedResident = false;
+        Uint64 promotedChangeSerial = 0;
     };
 
     // Supplies a command buffer that is recording and outside any render pass,
@@ -80,10 +102,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // Recreate all per-frame transient arenas
         Bool RecreateTransientArenas(Uint32 frameCount);
         void BeginFrame(Uint32 frameIndex);
-        // Drains every frame slot's deferred buffer/resource releases (and the
+        // Drains every frame slot's deferred buffer/resource releases. Only valid when
-        // transient arena's parked superseded blocks). Only valid when the
+        // the caller has proven every queue submission complete; used by the present-less
-        // caller has proven every queue submission complete; used by the
+        // frame-boundary drain. Deliberately does NOT touch the transient arena's parked
-        // present-less frame-boundary drain.
+        // superseded blocks: those are still named by this frame's slices (see the
+        // definition), and only a frame rewind retires them.
         void CollectAllDeferredReleases();
         // All previously submitted GPU work has completed (vkDeviceWaitIdle).
         void NotifyDeviceIdle();
@@ -121,6 +144,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         void OnResourceDestroyed(SharedPtr<MG_State::GLState::BackendBufferResource>&& resource);
 
         Uint64 GetFrameSerial() const { return m_frameSerial; }
+        // Highest value handed to any VkBufferResource::sliceEpoch. Unchanged since a
+        // memo was taken means no buffer this manager owns changed which slice it hands
+        // back, and none was persistently mapped, in between - so a memo of resolved
+        // slices needs no per-buffer re-check. See AcquirePersistentMap for the mapping half.
+        Uint64 GetSliceEpochCounter() const { return m_sliceEpochCounter; }
         // Highest frame serial whose GPU work is known complete; serials at or
         // below it may be considered signaled. Drives IsResourceBusy and the
         // backend GL fence objects.
@@ -147,6 +175,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         void DestroyAllDeferredReleases();
         void TrackLiveResource(const SharedPtr<VkBufferResource>& resource);
         void ReleaseAllLiveResources();
+        // See VkBufferResource::sliceEpoch.
+        void BumpSliceEpoch(VkBufferResource& resource) { resource.sliceEpoch = ++m_sliceEpochCounter; }
 
         VkBufferManagerInitInfo m_initInfo{};
         BufferArena m_transientUploadArena;
@@ -154,8 +184,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Vector<Vector<VkBufferObject>> m_deferredBufferReleases;
         Vector<Vector<SharedPtr<VkBufferResource>>> m_deferredResourceReleases;
         Vector<WeakPtr<VkBufferResource>> m_liveResources;
+    // Size m_liveResources had just after the last sweep; the next sweep waits for it to double.
+    SizeT m_liveResourcesLastPruned = 0;
         Uint32 m_currentFrameIndex = 0;
         Uint64 m_frameSerial = 1;
         Uint64 m_completedSerialFloor = 0;
+        // Never reset (not even by Shutdown): a value handed to a resource must stay
+        // unique for the process, or a memo taken before a re-initialize could match
+        // a different resource's state after it.
+        Uint64 m_sliceEpochCounter = 0;
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferObject.cpp

@@ -76,7 +76,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         const VkResult result =
             vmaCreateBuffer(m_allocator, &bufferInfo, &allocationInfo, &m_buffer, &m_allocation, nullptr);
         if (result != VK_SUCCESS) {
-            MGLOG_E("VkBufferObject::Create failed: vmaCreateBuffer returned %d", result);
+            MGLOG_E_ONCE("VkBufferObject::Create failed: vmaCreateBuffer returned %d", result);
             m_allocator = nullptr;
             m_buffer = VK_NULL_HANDLE;
             m_allocation = nullptr;
@@ -108,7 +108,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         const VkResult mapResult = vmaMapMemory(m_allocator, m_allocation, &m_mappedData);
         if (mapResult != VK_SUCCESS || m_mappedData == nullptr) {
-            MGLOG_E("VkBufferObject::Map failed: vmaMapMemory returned %d", mapResult);
+            MGLOG_E_ONCE("VkBufferObject::Map failed: vmaMapMemory returned %d", mapResult);
             m_mappedData = nullptr;
             return nullptr;
         }
@@ -138,14 +138,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         const Bool wasMapped = IsMapped();
         void* mapped = wasMapped ? m_mappedData : Map();
         if (mapped == nullptr) {
-            MGLOG_E("VkBufferObject::Upload failed: unable to map buffer");
+            MGLOG_E_ONCE("VkBufferObject::Upload failed: unable to map buffer");
             return false;
         }
 
         Memcpy(static_cast<Uint8*>(mapped) + offset, data, static_cast<SizeT>(size));
         const VkResult flushResult = vmaFlushAllocation(m_allocator, m_allocation, offset, size);
         if (flushResult != VK_SUCCESS) {
-            MGLOG_E("VkBufferObject::Upload failed: vmaFlushAllocation returned %d", flushResult);
+            MGLOG_E_ONCE("VkBufferObject::Upload failed: vmaFlushAllocation returned %d", flushResult);
             if (!wasMapped) {
                 Unmap();
             }
@@ -170,22 +170,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         const VkResult result = vmaInvalidateAllocation(m_allocator, m_allocation, offset, resolvedSize);
         if (result != VK_SUCCESS) {
-            MGLOG_E("VkBufferObject::Invalidate failed: vmaInvalidateAllocation returned %d", result);
+            MGLOG_E_ONCE("VkBufferObject::Invalidate failed: vmaInvalidateAllocation returned %d", result);
             return false;
         }
         return true;
     }
 
-    BufferSlice VkBufferObject::GetSlice(VkDeviceSize offset, VkDeviceSize size) const {
-        MOBILEGL_ASSERT(offset <= m_size, "VkBufferObject::GetSlice offset out of range");
-        const VkDeviceSize resolvedSize = (size == VK_WHOLE_SIZE) ? (m_size - offset) : size;
-        MOBILEGL_ASSERT(offset + resolvedSize <= m_size, "VkBufferObject::GetSlice range out of bounds");
-
-        BufferSlice slice{};
-        slice.buffer = m_buffer;
-        slice.offset = offset;
-        slice.size = resolvedSize;
-        slice.mapped = (m_mappedData != nullptr) ? static_cast<Uint8*>(m_mappedData) + offset : nullptr;
-        return slice;
-    }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferObject.h

@@ -48,7 +48,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         VkBuffer GetHandle() const { return m_buffer; }
         VkDeviceSize GetSize() const { return m_size; }
-        BufferSlice GetSlice(VkDeviceSize offset = 0, VkDeviceSize size = VK_WHOLE_SIZE) const;
+        // Inline: runs on the per-draw acquire path (a resident buffer bind is a
+        // GetSlice per binding), where an out-of-line call was measurable.
+        BufferSlice GetSlice(VkDeviceSize offset = 0, VkDeviceSize size = VK_WHOLE_SIZE) const {
+            MOBILEGL_ASSERT(offset <= m_size, "VkBufferObject::GetSlice offset out of range");
+            const VkDeviceSize resolvedSize = (size == VK_WHOLE_SIZE) ? (m_size - offset) : size;
+            MOBILEGL_ASSERT(offset + resolvedSize <= m_size, "VkBufferObject::GetSlice range out of bounds");
+
+            BufferSlice slice{};
+            slice.buffer = m_buffer;
+            slice.offset = offset;
+            slice.size = resolvedSize;
+            slice.mapped = (m_mappedData != nullptr) ? static_cast<Uint8*>(m_mappedData) + offset : nullptr;
+            return slice;
+        }
         void* GetMappedData() const { return m_mappedData; }
         Bool IsMapped() const { return m_mappedData != nullptr; }
         Bool IsValid() const { return m_allocator != nullptr && m_buffer != VK_NULL_HANDLE && m_allocation != nullptr; }

MobileGL/MG_Backend/DirectVulkan/Renderer/VkClearManager.cpp

@@ -166,7 +166,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     void VkClearManager::MergeClearPayload(ClearAttachmentPayload& dst, const ClearAttachmentPayload& src) {
         dst.mask |= src.mask;
         if ((src.mask & GL_COLOR_BUFFER_BIT) != 0) {
+            // The whole colour story travels together (same rule as
+            // VkRenderPassManager::QueueRenderbufferClear): a glClearBufferiv/uiv
+            // payload carries its value in colorInt/colorUint and its branch selector
+            // in colorEncoding - dropping them here would leave the pending clear
+            // reading as an all-zero float one.
             dst.color = src.color;
+            dst.colorEncoding = src.colorEncoding;
+            dst.colorInt = src.colorInt;
+            dst.colorUint = src.colorUint;
         }
         if ((src.mask & GL_DEPTH_BUFFER_BIT) != 0) {
             dst.depth = src.depth;

MobileGL/MG_Backend/DirectVulkan/Renderer/VkRenderPassManager.cpp

@@ -87,9 +87,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     static VkImageViewType ResolveAttachmentViewType(
         const MG_State::GLState::FramebufferAttachmentObject& attachment,
         const VkTextureManager::TextureResource& resource) {
-        return !attachment.IsLayered() && IsCubeMapFaceUploadTarget(attachment.GetTextureUploadTarget()) ?
+        if (attachment.IsLayered()) {
-            VK_IMAGE_VIEW_TYPE_2D :
+            return resource.viewType;
-            resource.viewType;
+        }
+        // A non-layered attachment names ONE layer, so the view over it is a plain 2D view whatever
+        // the image's own view type is. The cube-face upload targets always meant this; a cube map
+        // array attached through glFramebufferTextureLayer means it too, and a CUBE_ARRAY view over
+        // a single layer is not a legal attachment. The CUBE arm is inert today - no frontend path
+        // produces a non-layered cube attachment without a face upload target - and is kept for
+        // symmetry with CUBE_ARRAY.
+        if (IsCubeMapFaceUploadTarget(attachment.GetTextureUploadTarget()) ||
+            resource.viewType == VK_IMAGE_VIEW_TYPE_CUBE_ARRAY || resource.viewType == VK_IMAGE_VIEW_TYPE_CUBE) {
+            return VK_IMAGE_VIEW_TYPE_2D;
+        }
+        return resource.viewType;
     }
 
     static MG_State::GLState::ITextureObject* ResolveCompleteColorAttachmentTexture(
@@ -112,7 +123,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
 
         if (!attachment.IsComplete()) {
-            MGLOG_W("GetOrCreateRenderPass: draw buffer slot %u (%s) on FBO %u has an incomplete texture attachment; using VK_ATTACHMENT_UNUSED",
+            MGLOG_W_ONCE("GetOrCreateRenderPass: draw buffer slot %u (%s) on FBO %u has an incomplete texture attachment; using VK_ATTACHMENT_UNUSED",
                     drawBufferIndex,
                     MG_Util::ConvertFramebufferAttachmentTypeToString(attachmentType).c_str(),
                     fbo.GetExternalIndex());
@@ -121,7 +132,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         auto* texture = attachment.GetTexture().get();
         if (texture == nullptr) {
-            MGLOG_W("GetOrCreateRenderPass: draw buffer slot %u (%s) on FBO %u resolved to a null texture; using VK_ATTACHMENT_UNUSED",
+            MGLOG_W_ONCE("GetOrCreateRenderPass: draw buffer slot %u (%s) on FBO %u resolved to a null texture; using VK_ATTACHMENT_UNUSED",
                     drawBufferIndex,
                     MG_Util::ConvertFramebufferAttachmentTypeToString(attachmentType).c_str(),
                     fbo.GetExternalIndex());
@@ -300,7 +311,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         VkSampleCountFlagBits sampleCount = VK_SAMPLE_COUNT_1_BIT;
         if (!TryResolveSampleCountFlagBits(renderbuffer->GetSamples(), sampleCount)) {
-            MGLOG_E("GetOrCreateRenderbufferResource: unsupported renderbuffer sample count %d for renderbuffer %u",
+            MGLOG_E_ONCE("GetOrCreateRenderbufferResource: unsupported renderbuffer sample count %d for renderbuffer %u",
                     renderbuffer->GetSamples(),
                     renderbuffer->GetExternalIndex());
             return nullptr;
@@ -446,7 +457,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             m_physicalDevice, format, imageInfo.imageType, imageInfo.tiling, imageInfo.usage, imageInfo.flags,
             &imageFormatProperties);
         if (imageFormatResult != VK_SUCCESS || (imageFormatProperties.sampleCounts & sampleCount) == 0) {
-            MGLOG_E("GetOrCreateRenderbufferResource: unsupported renderbuffer format=%d samples=%d for renderbuffer %u",
+            MGLOG_E_ONCE("GetOrCreateRenderbufferResource: unsupported renderbuffer format=%d samples=%d for renderbuffer %u",
                     static_cast<Int>(format),
                     static_cast<Int>(sampleCount),
                     renderbuffer->GetExternalIndex());
@@ -820,6 +831,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // recreated since (texture + renderbuffer image epochs), and no pending clear (which alters
         // load ops). Any of these differing forces the full recompute below. Portable to VK 1.1.
         if (activeRenderPass != nullptr && m_rpFastValid && m_rpFastFbo == &fbo &&
+            m_rpFastFboLifetimeId == fbo.GetLifetimeId() &&
             m_rpFastFboVersion == fbo.GetObjectVersion() && m_rpFastSwapchainIndex == swapchainImageIndex &&
             m_rpFastTexEpoch == m_textureManager.GetTextureImageEpoch() &&
             m_rpFastRbEpoch == m_renderbufferImageEpoch &&
@@ -844,6 +856,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             // epochs AFTER ComputeHash: its attachment SyncTexture can create an image (bump the epoch).
             m_rpFastValid = true;
             m_rpFastFbo = &fbo;
+            m_rpFastFboLifetimeId = fbo.GetLifetimeId();
             m_rpFastFboVersion = fbo.GetObjectVersion();
             m_rpFastSwapchainIndex = swapchainImageIndex;
             m_rpFastTexEpoch = m_textureManager.GetTextureImageEpoch();
@@ -918,7 +931,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                     const auto& renderbuffer = rbAtt.GetRenderbuffer();
                     auto* rbResource = GetOrCreateRenderbufferResource(renderbuffer);
                     if (rbResource == nullptr || (rbResource->aspect & VK_IMAGE_ASPECT_COLOR_BIT) == 0) {
-                        MGLOG_E("GetOrCreateRenderPass: draw buffer slot %u on FBO %u has an unsupported color "
+                        MGLOG_E_ONCE("GetOrCreateRenderPass: draw buffer slot %u on FBO %u has an unsupported color "
                                 "renderbuffer %u; using VK_ATTACHMENT_UNUSED",
                                 i, fbo.GetExternalIndex(), renderbuffer->GetExternalIndex());
                         continue;
@@ -1094,7 +1107,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                     adoptRenderPassSampleCount(attachmentSampleCount, "color", texture->GetExternalIndex());
 
                     if (!hasClear && trackedColorLayout == VK_IMAGE_LAYOUT_UNDEFINED) {
-                        MGLOG_W("GetOrCreateRenderPass: color attachment textureId=%d starts with undefined layout and no clear; "
+                        MGLOG_W_ONCE("GetOrCreateRenderPass: color attachment textureId=%d starts with undefined layout and no clear; "
                                 "using LOAD_OP_DONT_CARE",
                                 texture->GetExternalIndex());
                         desc.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
@@ -1150,7 +1163,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             isUsableDepthStencilAttachment(depthAtt) && isUsableDepthStencilAttachment(stencilAtt) &&
             !sameDepthStencilAttachmentObject(depthAtt, stencilAtt);
         if (hasDistinctDepthAndStencilAttachments) {
-            MGLOG_E("GetOrCreateRenderPass: separate depth/stencil attachments are not supported yet; using the depth attachment and ignoring the standalone stencil attachment for framebuffer %u",
+            MGLOG_E_ONCE("GetOrCreateRenderPass: separate depth/stencil attachments are not supported yet; using the depth attachment and ignoring the standalone stencil attachment for framebuffer %u",
                     fbo.GetExternalIndex());
         }
         if (selectedDepthStencilAttachment != nullptr) {
@@ -1212,7 +1225,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             depthAttachmentDescription.finalLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
             depthAttachmentDescription.initialLayout = loadInfo.initialLayout;
             if (trackedDepthLayout == VK_IMAGE_LAYOUT_UNDEFINED && (!clearDepth || !clearStencil)) {
-                MGLOG_W("GetOrCreateRenderPass: depth/stencil attachment id=%d starts with undefined layout "
+                MGLOG_W_ONCE("GetOrCreateRenderPass: depth/stencil attachment id=%d starts with undefined layout "
                         "and partial/no clear; using DONT_CARE for uncleared aspects",
                         depthAttachmentId);
             }
@@ -1496,7 +1509,23 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             ClearAttachmentPayload clearPayload{};
             SharedPtr<MG_State::GLState::ITextureObject> liveTexture;
             if (pending.hasInlinePayload) {
+                // The inline payload was snapshotted when the entry was CREATED, but the
+                // clear VALUE is not part of the entry's hash - a cache hit with a newer
+                // glClear would replay the creation-time value and drop the new one (the
+                // texture path below is immune because it re-reads the live payload).
+                // Same defense as ClearAttachmentsOnActiveRenderPass: prefer the live
+                // pending clear, fall back to the snapshot only when none is queued.
+                if (s_renderPassManager != nullptr &&
+                    s_renderPassManager->GetPendingRenderbufferClear(pending.renderbuffer, clearPayload)) {
+                    if ((clearPayload.mask & GL_COLOR_BUFFER_BIT) != 0 && pending.renderbuffer != nullptr &&
+                        MG_Util::GetBaseInternalFormatComponentCount(pending.renderbuffer->GetInternalFormat()) ==
+                            3) {
+                        // RGB renderbuffers are backed by an RGBA image; the missing alpha reads as 1.
+                        ForceOpaqueClearAlpha(clearPayload);
+                    }
+                } else {
                     clearPayload = pending.inlinePayload;
+                }
             } else {
                 if (pending.key.texture == nullptr ||
                     !s_clearManager->GetPendingClear(pending.key, clearPayload, liveTexture)) {

MobileGL/MG_Backend/DirectVulkan/Renderer/VkRenderPassManager.h

@@ -16,6 +16,7 @@
 #include "MG_State/GLState/FramebufferState/FramebufferObject.h"
 
 #include <Includes.h>
+#include <unordered_map>
 #include <vk_mem_alloc.h>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
@@ -100,6 +101,42 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             std::swap(layers, that.layers);
             std::swap(lastUsedFrame, that.lastUsedFrame);
         }
+        // Move ASSIGNMENT, not just construction. The move constructor above and the
+        // destructor below each independently suppress the implicit one, which left the
+        // type move-constructible but not move-assignable - and therefore not swappable,
+        // which std::swap(pair&, pair&) requires. That was invisible while UnorderedMap
+        // only ever move-CONSTRUCTED an element into a fresh slot. ska::flat_hash_map
+        // probes robin-hood: inserting swaps the entry being placed against the one
+        // already sitting in the slot whenever it has travelled further from its desired
+        // position, so the mapped type has to be swappable or the table fails to
+        // instantiate at all.
+        //
+        // SWAP SEMANTICS, exactly like the move constructor: this does not release the
+        // destination's handles, it parks them in `that`, which destroys them when it
+        // dies. That is correct for the only caller - std::swap, whose temporary expires
+        // immediately - and it is what keeps the three-move sequence from destroying a
+        // live render pass. It is NOT correct for a hand-written `a = std::move(b)` where
+        // `a` held live handles and `b` outlives the statement: those handles would then
+        // survive until `b` dies. There is no such caller; add a destroy-then-steal
+        // assignment before writing one.
+        RenderPassEntry& operator=(RenderPassEntry&& that) noexcept {
+            if (this != &that) {
+                std::swap(hash, that.hash);
+                std::swap(renderPass, that.renderPass);
+                std::swap(framebuffer, that.framebuffer);
+                std::swap(compatibilityHash, that.compatibilityHash);
+                std::swap(pendingClearAttachments, that.pendingClearAttachments);
+                std::swap(trackedAttachmentLayouts, that.trackedAttachmentLayouts);
+                std::swap(attachmentCount, that.attachmentCount);
+                std::swap(colorAttachmentCount, that.colorAttachmentCount);
+                std::swap(hasDepthStencilAttachment, that.hasDepthStencilAttachment);
+                std::swap(sampleCount, that.sampleCount);
+                std::swap(extent, that.extent);
+                std::swap(layers, that.layers);
+                std::swap(lastUsedFrame, that.lastUsedFrame);
+            }
+            return *this;
+        }
         RenderPassEntry(
             Uint64 hash,
             VkRenderPass renderpass,
@@ -252,6 +289,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // or a pending clear. Portable to Vulkan 1.1 (no dynamic_rendering / imageless FB needed).
         Bool m_rpFastValid = false;
         const MG_State::GLState::FramebufferObject* m_rpFastFbo = nullptr;
+        // The FBO's never-reused lifetime id joins the raw pointer + Uint16 version:
+        // a deleted FBO reallocated at the same address whose fresh setup performed
+        // the same number of version bumps would otherwise compare equal (both count
+        // from 0), serving the dead framebuffer's pass to the new object.
+        Uint64 m_rpFastFboLifetimeId = 0;
         Uint16 m_rpFastFboVersion = 0;
         Uint32 m_rpFastSwapchainIndex = 0;
         Uint64 m_rpFastTexEpoch = 0;
@@ -314,7 +356,31 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Uint64 deferredAtFrame = 0;
         };
 
-        UnorderedMap<MG_State::GLState::RenderbufferObject*, RenderbufferResource> m_renderbufferResources;
+        // Node-based std::unordered_map, deliberately NOT the open-addressing UnorderedMap:
+        // callers cache a RenderbufferResource* - or a bare &resource->layout - and then make further
+        // calls that touch this map. BlitFramebuffer is the one that bit: it resolves the source and
+        // destination colour bindings (ResolveColorBlitBinding caches &rbResource->layout), then
+        // materializes the source's pending clear, which looks that same resource up again. Growing
+        // an open-addressed table relocates every element, so the cached pointer went on to name
+        // freed storage still holding the pre-clear VK_IMAGE_LAYOUT_UNDEFINED; BlitFramebuffer bailed
+        // out at "source image layout is undefined", silently dropping the blit -
+        // renderbuffers_storage_multisample read back zero instead of the clear colour on exactly the
+        // iterations that grew the table.
+        //
+        // Reordering the materialize ahead of the resolves - the fix ReadPixels got - does not cover
+        // this: the destination resolve still runs after the source pointer is taken. The depth blit,
+        // GetOrCreateRenderPass's depthRenderbufferResource and ReadDepthStencilPixels cache the same
+        // kind of pointer, so the invariant belongs in the container rather than in a per-call-site
+        // ordering rule. m_textureResources is node-based for the same reason.
+        //
+        // The case for keeping this node-based got STRONGER with ska::flat_hash_map, so do not read
+        // the paragraph above as merely historical: ska erases by shifting the rest of the probe
+        // cluster backwards into the hole, so erasing one renderbuffer relocates OTHER renderbuffers'
+        // entries - a cached pointer can now be invalidated by a key it has nothing to do with, which
+        // no call-site ordering rule can defend against. (What did change: ska's operator[] returns on
+        // a hit before it runs its grow check, so a plain lookup of a PRESENT key no longer relocates.
+        // That narrows the insert hazard; it does not touch the erase one.)
+        std::unordered_map<MG_State::GLState::RenderbufferObject*, RenderbufferResource> m_renderbufferResources;
         UnorderedMap<MG_State::GLState::RenderbufferObject*, PendingRenderbufferClear> m_pendingRenderbufferClears;
         Vector<DeferredRenderbufferRelease> m_deferredRenderbufferReleases;
         // Supported sample counts per attachment format, so per-draw resource lookups

MobileGL/MG_Backend/DirectVulkan/Renderer/VkTextureManager.h

@@ -59,6 +59,17 @@ public:
         // VK_KHR_image_format_list is enabled: MUTABLE_FORMAT images can name the exact set of
         // formats they will be viewed as, which is what lets a tiler keep them compressed.
         Bool imageFormatListSupported = false;
+        // Union of shader stages sampled-read barriers may name on this device; the renderer
+        // builds it from the enabled features because geometry/tessellation stage bits are
+        // invalid in a barrier when their feature is off.
+        VkPipelineStageFlags sampledReadStageMask = VK_PIPELINE_STAGE_VERTEX_SHADER_BIT |
+                                                    VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+                                                    VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
+        // Family of `graphicsQueue`; the manager creates its own command pool
+        // on it for the recycled upload-batch command buffers, so their parked
+        // allocations never sit in (and fragment) the renderer's shared pool
+        // that frame command buffers churn through every frame.
+        Uint32 graphicsQueueFamilyIndex = 0;
     };
 
     struct TextureResource {
@@ -299,9 +310,22 @@ public:
         static inline VmaAllocator s_allocator = VK_NULL_HANDLE;
     };
 
+    struct SampledTextureSnapshot {
+        VkImageView imageView = VK_NULL_HANDLE;
+        VkImageLayout layout = VK_IMAGE_LAYOUT_UNDEFINED;
+    };
+
     Bool Initialize(const InitInfo& initInfo);
     void Shutdown();
     void BeginFrame(Uint32 frameIndex);
+    // Submits the accumulated texture-upload batch (one command buffer, one
+    // vkQueueSubmit, one pooled fence) if any uploads are pending. MUST run
+    // before any other vkQueueSubmit on the shared graphics queue whose
+    // commands may consume an image the batch writes - the frame command
+    // buffer submit (mid-frame flush, readback, Present) and the
+    // preserve-on-recreate copy are the existing callers. No-op when the
+    // batch is empty.
+    void FlushPendingUploads();
     // Drains every frame slot's deferred image/view releases. Only valid when
     // the caller has proven every queue submission complete; used by the
     // present-less frame-boundary drain.
@@ -324,6 +348,13 @@ public:
                                                       VkImageLayout newLayout);
     Bool TransitionTextureForSampling(VkCommandBuffer commandBuffer, MG_State::GLState::ITextureObject& texture);
     Bool TransitionTextureForStorageImage(VkCommandBuffer commandBuffer, MG_State::GLState::ITextureObject& texture);
+    // Copies the complete sampler-visible mip range into a transient sampled image. The source is
+    // restored to its prior layout, so image-store descriptors continue to name the original image.
+    // The transient ownership is tied to the current frame slot and is safe through its submission.
+    Bool SnapshotTextureForSampling(VkCommandBuffer commandBuffer, MG_State::GLState::ITextureObject& texture,
+                                    SamplerNumericDomain numericDomain,
+                                    VkPipelineStageFlags consumerShaderStageMask,
+                                    SampledTextureSnapshot& outSnapshot);
 
     // Recording-generation bookkeeping for the pre-pass command stream. The
     // generation advances every time the frame command buffer (re)begins
@@ -360,17 +391,33 @@ public:
     // true - a false positive merely ends the render pass, a false negative would skip a barrier.
     Bool NeedsStorageImagePreparation(MG_State::GLState::ITextureObject& texture) const;
 
-    static VkImageAspectFlags ResolveSampledImageViewAspectMask(VkImageAspectFlags imageAspect);
+    // `depthStencilTextureMode` is the texture's GL_DEPTH_STENCIL_TEXTURE_MODE; it only decides
+    // anything for an image that carries both aspects. Defaulted so the call sites that have no
+    // texture in hand keep the depth-aspect answer they have always given.
+    static VkImageAspectFlags ResolveSampledImageViewAspectMask(VkImageAspectFlags imageAspect,
+                                                                GLenum depthStencilTextureMode = GL_DEPTH_COMPONENT);
     static VkFormat ResolveSampledImageViewFormat(VkFormat imageFormat, SamplerNumericDomain numericDomain);
     static Bool AreSampledImageViewFormatsCompatible(VkFormat imageFormat, VkFormat viewFormat);
     static Bool AreStorageImageViewFormatsCompatible(VkFormat imageFormat, VkFormat viewFormat);
 
+    // Moves `image` to `newLayout` and writes the new layout back through `trackedLayout`.
+    //
+    // The barrier covers EVERY array layer of the image, and there is deliberately no layer
+    // parameter to say otherwise: layout here is tracked per IMAGE (one `TextureResource::layout`,
+    // or one caller-owned variable), so a barrier narrower than the image would leave the layers it
+    // skipped in the old layout while the tracker claims they moved. Every transfer against a
+    // framebuffer attachment above layer 0 - glReadPixels, glBlitFramebuffer, glCopyTexSubImage,
+    // glCopyImageSubData - then ran its copy on a layer no barrier had transitioned.
+    //
+    // The mip range IS a parameter, because mip levels really are transitioned piecewise (see
+    // UpdateTrackedImageLayoutAfterAttachmentWrite and the mipmap generation loops): those callers
+    // move the complement of the level they wrote so the whole image converges on one layout again.
+    // Nothing does, or can, do that per layer.
     static Bool TransitionImageLayout(VkCommandBuffer commandBuffer, VkImage image, VkImageLayout& trackedLayout,
                                VkImageLayout newLayout, VkPipelineStageFlags srcStageMask,
                                VkPipelineStageFlags dstStageMask, VkAccessFlags srcAccessMask,
                                VkAccessFlags dstAccessMask, VkImageAspectFlags aspectMask,
-                               Uint32 baseMipLevel = 0, Uint32 levelCount = 1,
+                               Uint32 baseMipLevel = 0, Uint32 levelCount = 1);
-                               Uint32 layerCount = 1);
 
     SizeT CollectGarbage();
 
@@ -447,6 +494,9 @@ private:
     VkPhysicalDevice m_physicalDevice = VK_NULL_HANDLE;
     VmaAllocator m_allocator = nullptr;
     VkCommandPool m_commandPool = VK_NULL_HANDLE;
+    // Dedicated pool for the recycled upload-batch command buffers (see
+    // InitInfo::graphicsQueueFamilyIndex).
+    VkCommandPool m_uploadCommandPool = VK_NULL_HANDLE;
     VkQueue m_graphicsQueue = VK_NULL_HANDLE;
     Bool m_imageFormatListSupported = false;
     Uint32 m_currentFrameIndex = 0;
@@ -487,6 +537,10 @@ private:
     // Formats whose mutable-image probe failed on this device; their images are created
     // without MUTABLE_FORMAT_BIT so repeat syncs neither re-probe nor flag-mismatch.
     std::unordered_set<VkFormat> m_mutableFormatUnsupported;
+    // Formats whose 3D images refused VK_IMAGE_CREATE_2D_ARRAY_COMPATIBLE_BIT. Per format+usage,
+    // exactly like the mutable-format verdict above, so it is answered at image creation and
+    // remembered rather than probed once globally.
+    std::unordered_set<VkFormat> m_2dArrayCompatibleUnsupported;
     std::unordered_map<TextureIdentity, WeakPtr<MG_State::GLState::ITextureObject>, TextureIdentityHash> m_aliveObjects;
     std::unordered_map<TextureIdentity, TextureResource, TextureIdentityHash> m_textureResources;
     // Textures that have been bound to a GL image unit (see MarkStorageImageTexture).
@@ -496,15 +550,58 @@ private:
     std::unordered_map<VkFormat, VkSampleCountFlags> m_multisampleCountsByFormat;
     Vector<Vector<TextureResource>> m_deferredReleases;
     Vector<Vector<VkImageView>> m_deferredViewReleases;
+
+    // --- Batched upload machinery ---
+    // Uploads within a frame are recorded into ONE shared command buffer and
+    // submitted with ONE vkQueueSubmit at FlushPendingUploads (the renderer
+    // flushes before every frame-command-buffer submit). Staging memory comes
+    // from a pool of persistently-mapped, reusable blocks instead of a
+    // vmaCreateBuffer per upload.
+    struct UploadStagingBlock {
+        VkBuffer buffer = VK_NULL_HANDLE;
+        VmaAllocation allocation = nullptr;
+        Uint8* mapped = nullptr; // persistently mapped for the block's lifetime
+        VkDeviceSize capacity = 0;
+        VkDeviceSize cursor = 0; // bump cursor while the block backs the open batch
+    };
+    // Opens the batch command buffer lazily (allocates/reuses + begins recording).
+    VkCommandBuffer EnsureUploadBatchOpen();
+    // Bump-allocates `size` staging bytes for the open batch, growing onto a
+    // new/pooled block when the current one cannot fit. Returns the write
+    // pointer; outBuffer/outBaseOffset locate the space for copy commands.
+    Uint8* AcquireUploadStagingSpace(VkDeviceSize size, VkBuffer& outBuffer, VkDeviceSize& outBaseOffset);
+    void RecycleUploadStagingBlock(UploadStagingBlock&& block);
+    // Drops a recorded-but-unsubmitted batch on the floor. Shutdown only: the
+    // device is being torn down, so the lost texel data is unobservable.
+    void DiscardPendingUploadBatch();
+    void DestroyUploadPools();
+
+    Vector<UploadStagingBlock> m_freeUploadStagingBlocks;
+    VkDeviceSize m_freeUploadStagingBytes = 0;
+    Vector<VkCommandBuffer> m_freeUploadCommandBuffers;
+    Vector<VkFence> m_freeUploadFences;
+    Bool m_uploadBatchOpen = false;
+    VkCommandBuffer m_uploadBatchCommandBuffer = VK_NULL_HANDLE;
+    // Blocks whose staging bytes the open batch's copies reference (last =
+    // the block the bump cursor is currently allocating from).
+    Vector<UploadStagingBlock> m_uploadBatchBlocks;
+    // Images the open batch writes; consulted for the rare re-upload-after-
+    // draw flush and by DeferResourceRelease (an unsubmitted command buffer
+    // referencing a deferred-released image would escape every fence-based
+    // destruction proof, so the batch is flushed before the image is parked).
+    Vector<VkImage> m_uploadBatchImages;
+    VkDeviceSize m_uploadBatchStagingBytes = 0;
+
     // Texture uploads are submitted out-of-band but NOT waited on (waiting
     // behind the queue serialized the CPU against the previous frame's GPU
-    // work every time an animated atlas re-uploaded). Their transient objects
+    // work every time an animated atlas re-uploaded). Each flushed batch's
-    // are parked here and reclaimed once the upload fence signals.
+    // transients are parked here and RECYCLED (fence reset to the fence pool,
+    // command buffer reset to the CB pool, staging blocks back to the block
+    // pool) once the batch fence signals.
     struct PendingUploadReclaim {
         VkFence fence = VK_NULL_HANDLE;
         VkCommandBuffer commandBuffer = VK_NULL_HANDLE;
-        VkBuffer stagingBuffer = VK_NULL_HANDLE;
+        Vector<UploadStagingBlock> stagingBlocks;
-        VmaAllocation stagingAllocation = nullptr;
     };
     Vector<PendingUploadReclaim> m_pendingUploadReclaims;
 };

MobileGL/MG_Backend/DirectVulkan/Renderer/VkTimerQueryManager.cpp

@@ -15,7 +15,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         MOBILEGL_ASSERT(initInfo.device != VK_NULL_HANDLE, "VkTimerQueryManager::Initialize requires valid VkDevice");
         MOBILEGL_ASSERT(initInfo.frameCount > 0, "VkTimerQueryManager::Initialize requires non-zero frame count");
         if (initInfo.timestampValidBits == 0 || initInfo.timestampPeriodNs <= 0.0f || initInfo.slotsPerPool == 0) {
-            MGLOG_W("VkTimerQueryManager: timestamps unsupported (validBits=%u, period=%f, slots=%u)",
+            MGLOG_W_ONCE("VkTimerQueryManager: timestamps unsupported (validBits=%u, period=%f, slots=%u)",
                     initInfo.timestampValidBits, initInfo.timestampPeriodNs, initInfo.slotsPerPool);
             return false;
         }
@@ -35,7 +35,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         for (auto& poolState : m_pools) {
             const VkResult result = vkCreateQueryPool(m_device, &poolInfo, nullptr, &poolState.pool);
             if (result != VK_SUCCESS) {
-                MGLOG_E("VkTimerQueryManager: vkCreateQueryPool failed with %s", VkResultToString(result));
+                MGLOG_E_ONCE("VkTimerQueryManager: vkCreateQueryPool failed with %s", VkResultToString(result));
                 Shutdown();
                 return false;
             }
@@ -90,7 +90,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         auto& poolState = m_pools[frameIndex];
         if (poolState.cursor >= m_slotsPerPool) {
             if (!poolState.exhaustionWarned) {
-                MGLOG_W("VkTimerQueryManager: frame %u timestamp pool exhausted (%u slots); further timer queries "
+                MGLOG_W_ONCE("VkTimerQueryManager: frame %u timestamp pool exhausted (%u slots); further timer queries "
                         "this frame fall back to the frontend path",
                         frameIndex, m_slotsPerPool);
                 poolState.exhaustionWarned = true;
@@ -120,7 +120,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             m_device, m_pools[record.poolIndex].pool, record.slot, 1, sizeof(resultWithAvailability),
             resultWithAvailability, sizeof(Uint64), VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WITH_AVAILABILITY_BIT);
         if (result != VK_SUCCESS && result != VK_NOT_READY) {
-            MGLOG_E("VkTimerQueryManager: vkGetQueryPoolResults failed with %s", VkResultToString(result));
+            MGLOG_E_ONCE("VkTimerQueryManager: vkGetQueryPoolResults failed with %s", VkResultToString(result));
             return false;
         }
         if (resultWithAvailability[1] == 0) {

MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.h

@@ -211,10 +211,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                     GLsizei height, GLenum format, GLenum type, void* pixels);
         // Copy-and-repack core shared by depth-stencil ReadPixels and GetTexImage;
         // expects command recording to be active and any render pass already ended.
+        //
+        // `defaultFramebufferOrientation` is set only when the source is the swapchain's
+        // depth/stencil image, which this renderer stores display-side-up: the copy rect then
+        // has to be mapped out of GL's bottom-origin space and the copied rows re-oriented on
+        // the way back, exactly as the colour ReadPixels path does.
         void ReadDepthStencilImageToClient(VkImage image, VkFormat vkFormat, VkImageLayout* trackedLayout,
                                            VkImageAspectFlags imageAspect, Uint32 mipLevel, Uint32 baseArrayLayer,
                                            GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type,
-                                           void* pixels);
+                                           void* pixels, Bool defaultFramebufferOrientation = false);
         // Same-extent depth blit between images of different depth formats: host
         // round-trip with a per-texel re-encode (see BlitNamedFramebuffer).
         Bool BlitDepthAcrossFormats(FrameContext::FrameData& frame, VkImage srcImage, VkFormat srcFormat,
@@ -224,6 +229,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                     GLint dstY, GLint width, GLint height, VkImageLayout srcRestoreLayout,
                                     VkImageLayout dstRestoreLayout, Bool stencilAspect);
         static SizeT GetReadbackTexelSize(VkFormat sourceFormat);
+        // Map a GL bottom-left-origin rectangle into the display-oriented swapchain image.
+        // Quarter-turn surface transforms swap the copy extent's axes.
+        static Bool MapDefaultFramebufferReadbackRect(GLint x, GLint y, GLsizei width, GLsizei height,
+                                                      VkExtent2D imageExtent,
+                                                      VkSurfaceTransformFlagBitsKHR preTransform,
+                                                      VkOffset2D* imageOffset, VkExtent2D* imageCopyExtent);
+        // Reorder a tightly packed block copied with MapDefaultFramebufferReadbackRect back into
+        // GL row order. The input block has swapped dimensions for 90/270 degree transforms.
+        static Bool RemapDefaultFramebufferReadback(const Uint8* rawPixels, Uint32 logicalWidth,
+                                                    Uint32 logicalHeight,
+                                                    VkSurfaceTransformFlagBitsKHR preTransform,
+                                                    SizeT texelSize, Uint8* outPixels);
         static Bool ConvertReadbackPixels(const Uint8* sourcePixels, VkFormat sourceFormat,
                                           GLsizei width, GLsizei height, GLenum destinationFormat,
                                           GLenum destinationType, SizeT destinationRowStride,
@@ -293,6 +310,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // The samplerAnisotropy device feature was granted, so GL_TEXTURE_MAX_ANISOTROPY_EXT is
         // honored rather than accepted-and-ignored.
         Bool IsSamplerAnisotropySupported() const { return m_samplerAnisotropyFeatureEnabled; }
+        // ARB_base_instance extends indirect command records with a non-zero firstInstance and
+        // requires gl_InstanceID to remain zero-based. Vulkan needs both features to honor that
+        // complete contract: one legalizes the command word, the other enables the shader rebase.
+        Bool IsNonZeroIndirectBaseInstanceSupported() const {
+            return m_drawIndirectFirstInstanceFeatureEnabled && m_shaderDrawParametersFeatureEnabled;
+        }
         // Ensures the frame command buffer is recording (same lazy pattern as
         // SetupDraw) and writes a bottom-of-pipe timestamp into the current
         // frame's pool. Null when unsupported or the pool is exhausted.
@@ -327,6 +350,16 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Bool RecreateSwapchain();
 
     private:
+        // Tiered emission for an already-set-up multi-draw batch (state bound, index
+        // buffer bound for the indexed form). Tier 1: VK_EXT_multi_draw. Tier 2: one
+        // vkCmdDraw(Indexed)Indirect over a transient command array. Tier 3: unrolled
+        // vkCmdDraw(Indexed) loop. Tier eligibility is per-batch (uniform instance
+        // state for tier 1, firstInstance/feature legality for tier 2); every tier
+        // consumes the same param span, so contiguous-run merging done by the caller
+        // benefits all of them.
+        void EmitMultiDrawIndexed(VkCommandBuffer commandBuffer, const DrawIndexedCmdParam* pParams, Uint32 drawCount);
+        void EmitMultiDraw(VkCommandBuffer commandBuffer, const DrawCmdParam* pParams, Uint32 drawCount);
+
         struct BlitUniformData {
             float srcRect[4] = {0.f, 0.f, 1.f, 1.f};
             float dstRect[4] = {0.f, 0.f, 1.f, 1.f};
@@ -353,6 +386,31 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Uint32 samplerBinding = 0;
         };
 
+        // A single-sample staging image for multisample-resolve blits that also have to change
+        // orientation. vkCmdResolveImage cannot flip (it takes one offset per side, not the
+        // invertible pair vkCmdBlitImage takes), so a resolve into or out of the default
+        // framebuffer used to land the mirrored band. Resolving here first and then blitting from
+        // here separates the two operations, and each one then does only what it can express.
+        //
+        // Pooled rather than created per blit: the CTS runs hundreds of these back to back, and
+        // create-destroy per call would both cost allocations and, worse, need per-call deferred
+        // destruction to outlive the recording. It grows to the largest extent asked for and is
+        // reused; format changes recreate it.
+        struct MultisampleResolveScratchImage {
+            VkImage image = VK_NULL_HANDLE;
+            VmaAllocation allocation = VK_NULL_HANDLE;
+            VkFormat format = VK_FORMAT_UNDEFINED;
+            VkExtent2D extent = {0, 0};
+            VkImageLayout layout = VK_IMAGE_LAYOUT_UNDEFINED;
+        };
+        MultisampleResolveScratchImage m_msResolveScratch;
+        // Returns a scratch image at least `extent` in size with exactly `format`, transitioned to
+        // TRANSFER_DST and ready to be resolved into. Null image on failure (the caller then falls
+        // back to the direct resolve).
+        Bool AcquireMultisampleResolveScratchImage(VkCommandBuffer commandBuffer, VkFormat format,
+                                                   VkExtent2D extent);
+        void DestroyMultisampleResolveScratchImage();
+
         struct DeferredDepthMipmapCleanup {
             Vector<VkImageView> imageViews;
             Vector<VkFramebuffer> framebuffers;
@@ -435,15 +493,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         void* m_platformDisplay = nullptr;
         void* m_platformLibrary = nullptr;
         void* m_platformCloseDisplay = nullptr;
-        // Some real ICDs (e.g. NVIDIA's proprietary Linux driver) don't implement
+        // Whether the loader exposes VK_EXT_headless_surface, detected once in
-        // VK_EXT_headless_surface at all. Detected once in CreateInstance() from the
+        // CreateInstance() from the enumerated instance extensions. On desktop an
-        // enumerated instance extensions; when false, CreateSurface() falls back to a
+        // offscreen surface REQUIRES it: false is a clean, loud bring-up failure, never
-        // hidden Xlib window instead of vkCreateHeadlessSurfaceEXT.
+        // a substituted window. (Android is the one exception and has its own path -
+        // no Mali/Adreno driver seen so far exposes the extension, so a windowless
+        // context is given an AImageReader ANativeWindow that is never displayed.)
         Bool m_headlessSurfaceSupported = true;
-        // Set when CreateSurface() had to create its own Xlib window for the fallback
+        // Android has the same shortfall: no Mali/Adreno driver seen so far exposes
-        // above (rather than being handed one by the caller), so Shutdown() knows it
+        // VK_EXT_headless_surface, so a windowless (EGL pbuffer) context gets an
-        // owns that window and must destroy it.
+        // AImageReader's ANativeWindow to hand the WSI instead. Nothing is ever
-        Bool m_ownsFallbackXlibWindow = false;
+        // displayed - the reader's images are simply never acquired. Owned here, so
+        // Shutdown() deletes it.
+        void* m_fallbackImageReader = nullptr;
         VulkanRendererConfig m_config;
         Bool m_swapchainResizeRequested = false;
         // Presentation is suspended while the window is zero-area (minimized): the
@@ -471,10 +533,41 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Bool m_indexTypeUint8ExtensionEnabled = false;
         Bool m_logicOpFeatureEnabled = false;
         Bool m_multiDrawIndirectFeatureEnabled = false;
+        // drawIndirectFirstInstance gates indirect commands whose firstInstance != 0;
+        // cached at device creation because the tier-2 multi-draw path (a transient
+        // VkDrawIndexedIndirectCommand array) is illegal for such a sub-draw without it.
+        Bool m_drawIndirectFirstInstanceFeatureEnabled = false;
+        // VK_EXT_multi_draw: native batched submission for the CPU-side glMultiDraw*
+        // families (tier 1 of the multi-draw dispatch).
+        Bool m_multiDrawExtensionEnabled = false;
+        Uint32 m_maxMultiDrawCount = 0;
+        // Multi-draw dispatch tiers, resolved once at device creation from device support
+        // clamped by MOBILEGL_MAGMA_MULTIDRAW_MODE (a preference, never a demand):
+        //   tier 1 (ext):      one vkCmdDrawMulti(Indexed)EXT           - m_multiDrawAllowExt
+        //   tier 2 (indirect): one vkCmdDraw(Indexed)Indirect batch     - m_multiDrawAllowIndirect
+        //   tier 3 (unroll):   one vkCmdDraw(Indexed) per sub-draw      - always available
+        // m_multiDrawForceUnrollIndirect additionally forces the GPU-parameter
+        // glMultiDraw*Indirect paths onto their per-command loop (mode=unroll only).
+        Bool m_multiDrawAllowExt = false;
+        Bool m_multiDrawAllowIndirect = false;
+        Bool m_multiDrawForceUnrollIndirect = false;
         Bool m_samplerAnisotropyFeatureEnabled = false;
         Bool m_shaderDrawParametersExtensionEnabled = false;
         Bool m_shaderDrawParametersFeatureEnabled = false;
+        // Native subgroup topology, queried at device creation for the compute-module
+        // subgroup repairs (SubgroupSupportPolicy.h) and the REQUIRE_FULL_SUBGROUPS
+        // stage flag; 0 / false when the device has no usable compute subgroups or
+        // MOBILEGL_DISABLE_SUBGROUP forced them off.
+        Uint32 m_nativeSubgroupSize = 0;
+        Bool m_nativeSubgroupSupported = false;
+        Bool m_computeFullSubgroupsFeatureEnabled = false;
+        // VkPhysicalDeviceSubgroupSizeControlProperties::maxComputeWorkgroupSubgroups;
+        // 0 when the extension (and therefore the full-subgroups flag) is unavailable.
+        Uint32 m_maxComputeWorkgroupSubgroups = 0;
         Bool m_unformattedFloatStorageImagesEnabled = false;
+        // Set only after descriptor-indexing feature AND property queries prove that
+        // update-after-bind is legal for every descriptor category this renderer emits.
+        ProgramFactory::UpdateAfterBindLimits m_updateAfterBindLimits{};
         // fillModeNonSolid gates VK_POLYGON_MODE_LINE/_POINT (glPolygonMode); independentBlend gates
         // per-draw-buffer color write masks (glColorMaski). Both are cached at device creation and
         // drive a runtime fallback when the device lacks them.
@@ -485,6 +578,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // needs no feature). Both cached at device creation and drive a hard-fail-at-draw when absent.
         Bool m_dualSrcBlendFeatureEnabled = false;
         Bool m_primitiveTopologyListRestartFeatureEnabled = false;
+        // multiViewport gates rasterizing into more than one of ARB_viewport_array's 16 viewports
+        // (gl_ViewportIndex). m_maxRasterizableViewports is min(MAX_VIEWPORTS, device limit), or 1
+        // when the feature is off, and is the viewportCount a gl_ViewportIndex-writing pipeline
+        // declares - it is NOT what GL_MAX_VIEWPORTS reports, which is the frontend state width.
+        Bool m_multiViewportFeatureEnabled = false;
+        Uint32 m_maxRasterizableViewports = 1;
+        // Union of shader stages sampled-read barriers may name; built at device creation
+        // because geometry/tessellation stage bits are invalid in a barrier when their
+        // feature is off (VUID-vkCmdPipelineBarrier-srcStageMask-04090/-04091), and
+        // ALL_GRAPHICS would also serialize against non-shader stages.
+        VkPipelineStageFlags m_sampledReadStageMask = VK_PIPELINE_STAGE_VERTEX_SHADER_BIT |
+                                                      VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+                                                      VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
         // Cached at device creation from the graphics queue family properties
         // and device limits; drives timer-query support.
         Uint32 m_timestampValidBits = 0;
@@ -495,9 +601,37 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                                                  VkDeviceSize countBufferOffset, Uint32 maxDrawCount,
                                                                  Uint32 stride);
         static inline PFNDrawIndexedIndirectCountFunc s_vkCmdDrawIndexedIndirectCount = nullptr;
+        // VK_EXT_multi_draw entry points, loaded at device creation when the extension
+        // (and its multiDraw feature) is enabled; null otherwise.
+        static inline PFN_vkCmdDrawMultiEXT s_vkCmdDrawMultiEXT = nullptr;
+        static inline PFN_vkCmdDrawMultiIndexedEXT s_vkCmdDrawMultiIndexedEXT = nullptr;
 
         // VK_EXT_transform_feedback (GL transform feedback capture)
         Bool m_transformFeedbackFeatureEnabled = false;
+        // VK_EXT_provoking_vertex. Vulkan's built-in convention is "provoking vertex first"; GL's
+        // default is LAST_VERTEX_CONVENTION, and GL derives BOTH flat shading and the transform
+        // feedback vertex order from it. provokingVertexLast alone fixes flat shading and the
+        // input-assembler capture order and has no dependency on transform feedback; only
+        // transformFeedbackPreservesProvokingVertex does.
+        Bool m_provokingVertexLastEnabled = false;
+        // transformFeedbackPreservesProvokingVertex was actually enabled at device creation. Kept
+        // separate because it is the only thing that arms
+        // VUID-VkGraphicsPipelineCreateInfo-topology-04884, the rule that forbids a TRIANGLE_FAN
+        // pipeline from asking for LAST on a device that cannot preserve a fan's provoking vertex.
+        Bool m_provokingVertexXfbPreserveEnabled = false;
+        // provokingVertexModePerPipeline: when VK_FALSE every pipeline in one render pass instance
+        // must agree on the mode, so glProvokingVertex(GL_FIRST_VERTEX_CONVENTION) cannot be honoured
+        // per draw and every pipeline takes GL's default (LAST) instead.
+        Bool m_provokingVertexModePerPipeline = false;
+        // transformFeedbackPreservesTriangleFanProvokingVertex.
+        Bool m_provokingVertexFanPreserved = false;
+        // Per-pipeline provoking-vertex mode. capturesXfbFromGeometryStage must be a LINK-TIME
+        // property of the program, never the dynamic "is transform feedback active" flag: the
+        // 8-entry m_pipelineMemo and the SetupDrawSnapshot fast path key on programObj.hash and
+        // the pipeline-state value hash, neither of which moves when glBeginTransformFeedback is
+        // called, so a dynamic input here would hand back a stale VkPipeline.
+        VkProvokingVertexModeEXT SelectProvokingVertexMode(VkPrimitiveTopology topology,
+                                                          Bool capturesXfbFromGeometryStage) const;
         // VK_EXT_vertex_attribute_divisor: without it every non-zero glVertexAttribDivisor
         // behaves as 1, because that is all Vulkan's instance input rate can express.
         Bool m_vertexAttributeDivisorEnabled = false;
@@ -586,7 +720,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Uint64 programHash = 0;
             Uint64 vertexInputHash = 0;
             Uint64 renderPassHash = 0;
-            Uint renderStateVersion = 0;
+            // VALUE hash of the pipeline-relevant fixed-function state (see
+            // ComputePipelineStateHash), not the monotonic pipeline-state version:
+            // the version never repeats, so a per-draw GL_BLEND toggle would miss
+            // all entries forever even though the state alternates between two
+            // values the memo already holds.
+            Uint64 pipelineStateHash = 0;
             ProgramFactory::CompileOptionFlags transformFlags = {};
             VkPipeline pipeline = VK_NULL_HANDLE;
         };
@@ -594,11 +733,40 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         PipelineMemoEntry m_pipelineMemo[kPipelineMemoSize];
         Uint32 m_pipelineMemoCount = 0;
         Uint32 m_pipelineMemoNext = 0;
+        // Hash of every fixed-function GL state the pipeline payload reads that the
+        // memo key's other fields (mode / program / vertex input / render pass /
+        // transform flags) do not already pin down. Equal hash under an equal rest
+        // of key => byte-identical PipelineCreatePayload. Cached per pipeline-state
+        // version: the version is monotonic and bumps on every pipeline-state
+        // change, so an unchanged (version, colorAttachmentCount) proves the state
+        // bytes are unchanged and the hash can be reused without re-reading them.
+        Uint64 ComputePipelineStateHash(Uint32 colorAttachmentCount) const;
+        Uint m_pipelineStateHashVersion = 0;
+        Uint32 m_pipelineStateHashColorCount = 0;
+        Uint64 m_pipelineStateHash = 0;
+        Bool m_pipelineStateHashValid = false;
+        // GetShaderTransformFlags memo. NOT pure in the pre-transform alone: the
+        // function also reads whether the bound DRAW framebuffer is the default one
+        // (only the default framebuffer gets the Y-flip and rotation bits - an FBO
+        // pass renders unflipped). Keyed on BOTH inputs; missing the FBO bit shipped
+        // an upside-down default-framebuffer pass after any render-to-texture
+        // (minecraft-1.17-main-menu retrace, whole frame flipped).
+        VkSurfaceTransformFlagBitsKHR m_baseTransformFlagsPreTransform =
+            VK_SURFACE_TRANSFORM_FLAG_BITS_MAX_ENUM_KHR;
+        Bool m_baseTransformFlagsIsDefaultFbo = false;
+        Bool m_baseTransformFlagsKeyValid = false;
+        Uint32 m_baseTransformFlagsCache = 0;
+        // isDefaultFbo must be the default-ness of the CURRENTLY bound draw framebuffer;
+        // every caller already has it in hand from its own guards.
+        Uint32 GetBaseTransformFlagsRaw(Bool isDefaultFbo);
         // Drops every memoized pipeline handle. Required at command-buffer
-        // boundaries and whenever any pipeline may have been destroyed.
+        // boundaries and whenever any pipeline may have been destroyed. Also drops
+        // the cached pipeline-state hash: the same boundaries can retire the GL
+        // context whose monotonic version the cache is keyed on.
         void InvalidatePipelineMemo() {
             m_pipelineMemoCount = 0;
             m_pipelineMemoNext = 0;
+            m_pipelineStateHashValid = false;
         }
         UnorderedMap<ProgramFactory::HashType, VkPipeline> m_computePipelines;
         UniquePtr<ProgramFactory> m_programFactory;
@@ -639,9 +807,26 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Uint32 m_lastLodProgramVersion = 0;
         Uint64 m_lastLodBindGeneration = 0;
         Uint64 m_lastLodParamsSum = 0;
+        // Sampling-resolution generation at probe time. The probe reads the effective
+        // sampler's filters/aniso/LOD range, whose setters bump only this counter -
+        // the params-version sum above never moves for them.
+        Uint64 m_lastLodSamplingGeneration = 0;
         ProgramFactory::CompileOptionFlags m_lastLodBaseFlags = {};
         ProgramFactory::CompileOptionFlags m_lastLodResultFlags = {};
 
+        // Does the current program's vertex stage declare the BaseVertex builtin? A property
+        // of the program's SPIR-V, so (lifetime id, backend-state version) is the whole key.
+        //
+        // Memoized rather than re-asked because asking means resolving the UN-zeroed program
+        // variant, and a program that only ever draws non-indexed would then compile a variant
+        // no draw uses AND re-stamp its use every draw, so the idle sweep could never retire
+        // it. With the memo the answer is known before the first lookup and only the variant
+        // the draw actually needs is resolved.
+        Bool m_lastBaseVertexQueryValid = false;
+        Uint64 m_lastBaseVertexProgramLifetimeId = 0;
+        Uint32 m_lastBaseVertexProgramVersion = 0;
+        Bool m_lastBaseVertexReads = false;
+
         // Snapshot behind TrySetupDrawFastPath. Values only: the program and
         // render-pass caches are open-addressing maps whose entries move on
         // insert, so no pointers into them are cached; the pipeline handle is
@@ -655,8 +840,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Uint64 programLifetimeId = 0;
             Uint32 programVersion = 0;
             const void* vao = nullptr;
+            // Same rule as VaoDrawMemo::vaoLifetimeId: (address, config version) is not an
+            // identity, because a recycled address can arrive carrying a config version
+            // the dead VAO also had (two mutations to configure one attribute is the
+            // common shape), and "the VAO did not move" would then skip the layout
+            // re-resolve for a different VAO.
+            Uint64 vaoLifetimeId = 0;
             Uint32 vaoConfigVersion = 0;
             const void* drawFbo = nullptr;
+            // Never-reused lifetime id beside the raw pointer + Uint16 version: a
+            // deleted FBO recycled at the same address with the same fresh version
+            // count would otherwise compare equal (same ABA as the render-pass
+            // manager's fast-path memo).
+            Uint64 drawFboLifetimeId = 0;
             Uint16 fboVersion = 0;
             Bool drawFboIsDefault = false;
             Uint renderStateVersion = 0;
@@ -670,20 +866,90 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Uint64 renderbufferImageEpoch = 0;
             Uint64 sampledContentSum = 0;
             Uint64 sampledParamsSum = 0;
+            // Guards the sampler-descriptor reuse hint: bumped by any sampler-object
+            // parameter or texture shape change (see GetSamplingResolutionGeneration),
+            // none of which the sums above cover.
+            Uint64 samplingResolutionGeneration = 0;
+            // Render-pass flavor input (DepthTest || StencilTest at snapshot time).
+            // A pipeline-state change that leaves this equal cannot change which
+            // render pass GetOrCreateRenderPass would pick, so the fast path may
+            // re-resolve just the pipeline against the active pass; a change that
+            // flips it must fall back to the full path's pass selection.
+            Bool drawUsesDepthStencil = false;
+            // The snapshotting draw's pipeline viewportCount. A pure function of the PROGRAM
+            // (writesViewportIndexBuiltin) and of a device feature fixed at renderer init, both
+            // of which the programLifetimeId/programVersion guards above already pin - carried
+            // here so the fast path does not re-fetch the program object to re-derive it.
+            Uint32 viewportCount = 1;
             IntVec2 renderPassExtent = {0, 0};
+            // colorAttachmentCount of the snapshotting draw's render pass: the
+            // pipeline-state hash input, so the fast path can refresh that hash and
+            // probe the pipeline memo after a state change without re-fetching the
+            // render-pass entry (the pass itself is pinned by renderPassHash above).
+            Uint32 renderPassColorCount = 0;
             VkPipeline pipeline = VK_NULL_HANDLE;
+            // layoutHash of the snapshotting draw's vertex-input state. The pipeline and
+            // the vertex-input pre-flight depend on the VAO only through this (plus the
+            // program, pinned separately), so a changed VAO whose aux memo carries the
+            // same layoutHash re-uses the snapshot's pipeline and pre-flight verdict
+            // outright - the VAO-cycling case Minecraft chunk rendering hits every draw.
+            Uint64 vaoLayoutHash = 0;
+            // Memoised ProgramFactory entry of the snapshotting draw, valid while
+            // (programLifetimeId, programVersion, resolvedTransformFlags) match - all
+            // checked above - AND the factory's cache structure epoch is unchanged (the
+            // cache is open-addressing and holds entries by value, so any insert/erase
+            // moves them). The fast path must re-stamp use through StampProgramUse when
+            // it bypasses GetOrCreateProgram, or the idle sweep could evict a live entry.
+            const ProgramFactory::VkProgramObject* programObj = nullptr;
+            Uint64 programFactoryEpoch = 0;
+            // Per-entry copies of the snapshotting draw's sampled set (the scratch
+            // vectors below hold only the LAST full-path draw's set, which with more
+            // than one snapshot entry is not necessarily this entry's program).
+            // sampledTextures/sampledResources carry the same epoch-guarded pointer
+            // lifetime rules as the scratch originals: textureEraseEpoch (checked
+            // every probe) declines the entry before any erased resource pointer
+            // could be dereferenced. sampledLayouts is the layout VALUE each
+            // resource held when this entry's descriptors were built (the
+            // descriptor-reuse hint needs the SAME layout, not just a sampleable
+            // one), and sampledBindingRecords feeds SampledBindingsUnchanged when
+            // the bind generation moved.
+            Vector<MG_State::GLState::ITextureObject*> sampledTextures;
+            Vector<VkTextureManager::TextureResource*> sampledResources;
+            Vector<VkImageLayout> sampledLayouts;
+            Vector<UniformManager::SampledBindingRecord> sampledBindingRecords;
         };
-        SetupDrawSnapshot m_setupDrawSnapshot;
+        // Program-keyed snapshot entries: program ping-pong (Sodium switches programs
+        // mid-frame every few draws) would otherwise evict the single snapshot on
+        // every switch and send every draw through the full path. Entries are found
+        // by programLifetimeId (MRU-first probe); every other guard stays per-probe,
+        // so a stale entry declines itself exactly like the old single snapshot did.
+        static constexpr Uint32 kSetupDrawSnapshotCount = 4;
+        SetupDrawSnapshot m_setupDrawSnapshots[kSetupDrawSnapshotCount];
+        Uint32 m_setupDrawSnapshotMru = 0;    // last entry that hit or was filled
+        Uint32 m_setupDrawSnapshotVictim = 0; // round-robin fill cursor when all entries are live
+        void InvalidateSetupDrawSnapshots() {
+            for (auto& snapshot : m_setupDrawSnapshots) {
+                snapshot.valid = false;
+            }
+        }
 
         // Per-draw scratch buffers (clear keeps capacity) — these paths run for every
         // draw call and must not allocate.
         Vector<MG_State::GLState::ITextureObject*> m_sampledTexturesScratch;
+        // Per-binding (texture, effective sampler) lifetime-id records from the same
+        // CollectSampledTextures walk that filled m_sampledTexturesScratch. The fast
+        // path shadow-compares against them (SampledBindingsUnchanged) when the
+        // texture bind generation moved, so a redundant glBindSampler/glBindTexture
+        // storm that resolves to the same bindings keeps the fast path.
+        Vector<UniformManager::SampledBindingRecord> m_sampledBindingRecordsScratch;
         // Parallel to m_sampledTexturesScratch, refilled by every SetupDraw's
         // first sampled-texture loop: the resolved backend resources, so the
         // post-transition loop can skip re-resolving textures whose layout is
         // already sampleable.
         Vector<VkTextureManager::TextureResource*> m_sampledResourcesScratch;
         Vector<MG_State::GLState::ITextureObject*> m_storageImageTexturesScratch;
+        Vector<UniformManager::SamplerImageFeedbackBinding> m_samplerImageFeedbackScratch;
+        Vector<UniformManager::SamplerBindingOverride> m_samplerImageBindingOverridesScratch;
         Vector<VkBuffer> m_vertexBuffersScratch;
         Vector<VkDeviceSize> m_vertexOffsetsScratch;
         Vector<VkVertexInputAttributeDescription> m_patchedAttributesScratch;
@@ -742,6 +1008,146 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         UnorderedMap<ConvertedVertexStreamKey, ConvertedVertexStream, ConvertedVertexStreamKeyHash>
             m_convertedVertexStreams;
 
+        // One VAO's resolved vkCmdBindVertexBuffers arguments, reusable by a later draw
+        // that would resolve them to the same thing. Consecutive draws in a chunk-renderer
+        // frame keep the program and the vertex layout and only swap the VAO, so a
+        // per-VAO memo turns the second and later draws through each VAO into a validate
+        // plus (usually skipped) rebind.
+        //
+        // Only whole-buffer bindings are memoised. Client-memory and format-converted
+        // streams re-upload from a range that depends on the draw's own vertex/index
+        // range, and synthetic bindings carry glVertexAttrib* values that are not part
+        // of any key here; a layout using any of them is never stored.
+        // Field order is hit-path cache locality, hot to cold: the per-draw validate
+        // reads the scalars and the EBO memo head, then only the first bindingCount
+        // elements of vkBuffers/vkOffsets; the per-binding revalidation arrays at the
+        // tail are touched once per frame at most.
+        struct ResolvedVertexBindings {
+            // Must equal DynamicStateShadow::kMaxShadowedVertexBindings (static_assert in
+            // the .cpp): past that width the bind shadow cannot skip a redundant bind
+            // either, so a wider layout resolves per draw. Minecraft-shaped layouts use four.
+            static constexpr Uint32 kMaxBindings = 8;
+
+            // Frame serial of the last completed resolve OR cross-frame revalidation.
+            // Zero until a resolve completes, and reset to zero before one starts, so a
+            // resolve that bails out midway cannot leave a half-filled entry matchable.
+            // Unlike the original frame-scoped memo, an entry whose buffers are all
+            // resident and unmapped is revalidated across frames (per-binding slice
+            // epoch compares) instead of re-resolved - see TryBindResolvedVertexBindings.
+            Uint64 frameSerial = 0;
+            // Identity of the resolved Vulkan layout: the VAO's content hash
+            // (VertexInputStateFactory::GetOrComputeHash - the same value the factory
+            // keys its entries on) fixes bindings.size(), each binding's base offset,
+            // which bindings are client/converted, and (through the mixed-in buffer
+            // addresses) which buffer each binding reads. Compared against the VAO's
+            // own hash memo on the hit path, so a hit never touches the factory entry.
+            VertexInputStateFactory::HashType vertexInputHash = 0;
+            // The program's vertex input layout: decides the synthetic-binding set and
+            // hence the total binding count.
+            Uint32 activeAttribMask = 0;
+            Uint32 bindingCount = 0;
+            // VkBufferManager::GetSliceEpochCounter() at resolve time. Still equal means
+            // no buffer anywhere changed its slice or was persistently mapped since, which
+            // settles every per-binding question below in one compare.
+            Uint64 sliceEpochCounter = 0;
+            // Any bound buffer already carrying a host map when the slice was resolved.
+            // Such a buffer can mutate its shadow with no API call, so it has to be
+            // re-pushed per draw and the one-compare path above cannot apply.
+            Bool anyBufferMapped = true;
+
+            // Resident element-buffer slice memo (skips the per-draw AcquireResidentSlice
+            // for the VAO's EBO, which cold-chases 500+ distinct resources in a
+            // chunk-cycling frame). Self-validating exactly like the bindings above: a hit
+            // requires the LIVE bound EBO pointer to equal indexBuffer AND either an
+            // unmoved manager-wide slice-epoch counter (nothing anywhere changed slices
+            // or gained a host map, the same one-compare rescue the vertex half uses) or
+            // that buffer's resource still carrying indexSliceEpoch (epochs are minted
+            // from a process-lifetime counter, so a recycled address can never
+            // revalidate). Restart-substituted and streamed EBOs are never stored.
+            // indexFrameSerial tracks the last frame the resource's GPU-use serial was
+            // stamped through this memo; 0 means no index memo. Independent of the
+            // vertex half: both are (pointer, epoch)-validated, so neither can serve
+            // stale state for the other.
+            const MG_State::GLState::BufferObject* indexBuffer = nullptr;
+            Uint64 indexSliceEpoch = 0;
+            // GetSliceEpochCounter() when the resource's epoch was last verified; only
+            // meaningful while indexFrameSerial matches the current frame serial.
+            Uint64 indexSliceEpochCounter = 0;
+            VkBuffer indexVkBuffer = VK_NULL_HANDLE;
+            VkDeviceSize indexSliceOffset = 0;
+            Uint64 indexFrameSerial = 0;
+            // The EBO carried a host map when the slice was recorded - the mirror of
+            // anyBufferMapped on the vertex half. A shadow-backed (non-adopted)
+            // persistent map mutates its shadow with no API call and no epoch bump, so
+            // the one-compare rescue must decline and re-run the acquire, whose
+            // SyncPersistentMappedRange is the push-down. A map taken AFTER the record
+            // is already covered: AcquirePersistentMap bumps the slice epoch for the
+            // request itself, adopted or declined.
+            Bool indexBufferMapped = false;
+
+            // Bound per draw (first bindingCount elements).
+            VkBuffer vkBuffers[kMaxBindings] = {};
+            VkDeviceSize vkOffsets[kMaxBindings] = {};
+            // Per binding: the VAO attribute location its buffer comes from, that buffer,
+            // and the buffer's VkBufferManager slice epoch when the slice was resolved.
+            // Only read by the per-frame revalidation and the something-moved fallback.
+            Uint8 attributeLocations[kMaxBindings] = {};
+            const MG_State::GLState::BufferObject* buffers[kMaxBindings] = {};
+            Uint64 sliceEpochs[kMaxBindings] = {};
+        };
+        // One direct-mapped slot of the per-VAO draw-memo table below. A slot belongs to
+        // the object whose (vaoKey, vaoLifetimeId) pair it carries: the address alone
+        // only picks the slot, and the never-reused lifetime id is what proves the slot
+        // is THIS VAO's, so the successor allocated onto a destroyed VAO's address
+        // always misses. That identity check is load-bearing and the content-hash
+        // validations below do NOT stand in for it - a recycled address under a
+        // byte-identical configuration reproduces the content hash exactly, which is
+        // how a destroyed VAO's resolved bindings were once handed to its successor's
+        // draw. The slot is still never dereferenced through vaoKey, and every fact it
+        // carries is still validated against live state before use:
+        //  - layoutHash/layoutAuxMasks are valid only while contentHash equals the LIVE
+        //    VAO's own hash memo (which the VAO's config version guards), so a config
+        //    change or a buffer rebind misses even for the same object.
+        //  - bindings revalidates per draw exactly as before (frame serial, content
+        //    hash, per-binding live buffer pointers and slice epochs).
+        struct alignas(64) VaoDrawMemo {
+            const MG_State::GLState::VertexArrayObject* vaoKey = nullptr;
+            // The VAO's never-reused lifetime id, checked alongside vaoKey. The pointer
+            // ALONE is not an identity: a deleted VAO's heap address is handed straight
+            // back by the next glGenVertexArrays-shaped allocation, and the successor then
+            // matched this slot and inherited the dead object's memos. Both stated
+            // defences failed with it, because both reduce to the content hash and the
+            // content hash's buffer-identity component was itself a recycled heap address.
+            Uint64 vaoLifetimeId = 0;
+            // The VAO content hash (VertexInputStateFactory::GetOrComputeHash) the two
+            // layout facts below were derived from; 0 while nothing valid is stored.
+            Uint64 contentHash = 0;
+            Bool layoutFactsValid = false;
+            // The resolved layout identity + packed (unsupported, location) masks -
+            // the exact values GetBackendAuxMemo used to serve, moved here so the
+            // per-draw probe stays inside this table's one hot line instead of
+            // touching a second cold line of every cycled VAO object.
+            Uint64 layoutHash = 0;
+            Uint64 layoutAuxMasks = 0;
+            ResolvedVertexBindings bindings;
+        };
+        // Fixed-size, allocated on first use, never rehashed or swept: entries are
+        // recycled in place on slot collisions (two-slot probe, older frame serial
+        // evicted), and stale entries self-invalidate through the compares above. A
+        // fixed table also makes every VaoDrawMemo/ResolvedVertexBindings pointer
+        // stable for the duration of a draw, which the EBO memo handoff
+        // (m_currentDrawResolvedEntry) relies on.
+        static constexpr Uint32 kVaoDrawMemoSlotCount = 2048; // power of two
+        Vector<VaoDrawMemo> m_vaoDrawMemoTable;
+        // Finds the slot holding `vao`, or recycles the older of its two candidate
+        // slots into an empty memo keyed on `vao`. Never returns null.
+        VaoDrawMemo* LookupVaoDrawMemo(const MG_State::GLState::VertexArrayObject* vao);
+        // The current draw's memo entry, set by UploadAndBindVertexBuffers and consumed
+        // by the same draw's UploadAndBindIndexBuffer (the EBO memo lives in the same
+        // entry). Valid ONLY within that window: the next draw's lookup can recycle the
+        // slot. Null when the draw's layout is not memoisable.
+        ResolvedVertexBindings* m_currentDrawResolvedEntry = nullptr;
+
         void CreateInstance();
         VkResult SetupDebugMessenger();
         VkResult DestroyDebugMessenger();
@@ -771,11 +1177,49 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             FrameContext::FrameData& frame,
             const MG_State::GLState::ProgramObject& program,
             const ProgramFactory::VkProgramObject& programObj);
+        // Vulkan forbids a sampled descriptor and writable storage descriptor from naming the
+        // same image subresource in one shader operation. Snapshot only the sampler side; the
+        // storage descriptor continues to name the application texture.
+        Bool PrepareSamplerImageFeedbackSnapshots(
+            FrameContext::FrameData& frame,
+            const MG_State::GLState::ProgramObject& program,
+            const ProgramFactory::VkProgramObject& programObj,
+            VkPipelineStageFlags consumerShaderStageMask);
+
+        // The per-draw dynamic-state tail (viewport, scissor, blend constants, depth
+        // bias, line width, stencil), gated behind one render-state-parameters-version
+        // compare per command buffer - see the gate fields in DynamicStateShadow.
+        // viewportCount is the bound pipeline's declared viewport count: 1 for every program that
+        // does not write gl_ViewportIndex (the memoized fast path), otherwise the renderer's
+        // rasterizable viewport count, which takes the unmemoized array path.
+        void ApplyDynamicDrawStateTail(FrameContext::FrameData& frame, const IntVec2& extent, Bool isDefaultFbo,
+                                       Uint32 viewportCount = 1);
+        void ApplyMultiViewportDynamicState(VkCommandBuffer commandBuffer, Uint32 viewportCount, const IntVec2& extent,
+                                            VkSurfaceTransformFlagBitsKHR preTransform, Bool isDefaultFbo);
+        VkRect2D ComputeGLScissorRect(Uint32 index, const IntVec2& extent,
+                                      VkSurfaceTransformFlagBitsKHR preTransform, Bool isDefaultFbo) const;
+        // How many viewports a draw with this program rasterizes into: 1 unless the program
+        // assigns gl_ViewportIndex AND the device enabled multiViewport. Both the pipeline's
+        // baked viewportCount and the dynamic arrays come from this one answer, so they cannot
+        // disagree.
+        Uint32 ResolveDrawViewportCount(Bool programWritesViewportIndex) const {
+            return programWritesViewportIndex && m_multiViewportFeatureEnabled ? m_maxRasterizableViewports : 1u;
+        }
 
         Bool UploadAndBindVertexBuffers(VkCommandBuffer commandBuffer, const MG_State::GLState::VertexArrayObject& vao,
                                         const ProgramFactory::VkProgramObject& programObj,
                                         const DrawCmdParam& drawParams,
                                         const IndexBufferView* pIndexBufferView);
+        // Binds `entry`'s memoised buffers when every input it was resolved from is
+        // still live and unchanged, else returns false and leaves nothing bound.
+        // vaoContentHash is the VAO's memoised content hash (GetBackendHashMemo), which
+        // pins the layout AND the bound buffers without resolving the factory entry.
+        // Non-const entry: a cross-frame revalidation refreshes its serial/epoch stamps.
+        Bool TryBindResolvedVertexBindings(VkCommandBuffer commandBuffer,
+                                           const MG_State::GLState::VertexArrayObject& vao,
+                                           ResolvedVertexBindings& entry,
+                                           Uint64 vaoContentHash,
+                                           Uint32 activeAttribMask, Uint64 frameSerial);
         Bool UploadAndBindIndexBuffer(FrameContext::FrameData& frame,
                                      const MG_State::GLState::VertexArrayObject& vao,
                                       const IndexBufferView* pIndexBufferView = nullptr);
@@ -791,11 +1235,28 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                                    GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
                                                    GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
                                                    GLenum filter);
+        // Clears one z slice of a VK_IMAGE_TYPE_3D colour image. See the call site in
+        // MaterializePendingClearForTexture for why a transfer clear cannot do this.
+        Bool ClearDepthSliceWithRenderPass(VkCommandBuffer commandBuffer,
+                                           MG_State::GLState::ITextureObject& texture, Uint32 mipLevel,
+                                           Uint32 depthSlice, const VkClearValue& clearValue);
         Bool MaterializePendingClearForTexture(VkCommandBuffer commandBuffer,
                                                MG_State::GLState::ITextureObject& texture);
         Bool MaterializePendingClearForRenderbuffer(
             VkCommandBuffer commandBuffer,
             const SharedPtr<MG_State::GLState::RenderbufferObject>& renderbuffer);
+        // The default framebuffer's twin of the two above. It cannot go through
+        // MaterializePendingClearForTexture: the default FBO's colour attachment is a
+        // placeholder texture object, and syncing THAT would clear a texture image nobody
+        // presents instead of the acquired swapchain image.
+        Bool MaterializePendingClearForDefaultFramebuffer(VkCommandBuffer commandBuffer,
+                                                          MG_State::GLState::FramebufferObject& fbo,
+                                                          FramebufferAttachmentType attachmentType);
+        // Its depth/stencil half: a different image (the swapchain's depth/stencil twin), a
+        // different clear command and per-aspect masking.
+        Bool MaterializePendingDepthStencilClearForDefaultFramebuffer(
+            VkCommandBuffer commandBuffer, const MG_State::GLState::FramebufferAttachmentObject& attachment,
+            const ClearAttachmentPayload& payload);
         VkPipeline GetOrCreateBlitPipeline(const RenderPassEntry& renderPassEntry);
         Bool GenerateDepthMipmapWithShader(FrameContext::FrameData& frame,
                                            MG_State::GLState::ITextureObject& texture,

MobileGL/MG_Backend/DirectVulkan/SubgroupSupportPolicy.h

@@ -0,0 +1,63 @@
+// MobileGL - MobileGL/MG_Backend/DirectVulkan/SubgroupSupportPolicy.h
+// Copyright (c) 2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#pragma once
+
+#include <Config.h>
+#include <Includes.h>
+
+namespace MobileGL::MG_Backend::DirectVulkan {
+    // The single decision point for how DirectVulkan implements GL_KHR_shader_subgroup,
+    // shared by capability advertisement (BackendObject) and module lowering
+    // (VulkanRenderer / ProgramFactory) so the two can never disagree.
+    //
+    // Native subgroups are the implementation whenever the device has them, whatever
+    // their width - subgroup operations execute on the hardware paths they were made
+    // for. Module-level repairs keep the GL contract intact around them:
+    //   - FixIterationRPSubgroupScratchPass patches the one known pack bug: iterationRP's
+    //     prefixSumCache[32], under-declared for sub-16-lane devices (8-lane lavapipe);
+    //   - FixIterationRPBarrierPass repairs Program 203's race between two reductions
+    //     reusing that scratch, when explicitly enabled;
+    //   - DeriveNumSubgroupsPass replaces the one builtin drivers get wrong
+    //     (gl_NumSubgroups) with the value the rest of the topology implies.
+    // The 32-lane shared-memory emulation (EmulateSubgroupsPass) is a LAST RESORT for
+    // devices with no subgroup support at all, and only when the user opts in with
+    // MOBILEGL_MAGMA_EMULATE_SUBGROUP=1; it never replaces available native operations.
+
+    inline constexpr Uint32 kEmulatedSubgroupSize = 32u;
+    inline constexpr Uint32 kEmulatedSubgroupStages = GL_COMPUTE_SHADER_BIT;
+    inline constexpr Uint32 kEmulatedSubgroupFeatures =
+        GL_SUBGROUP_FEATURE_BASIC_BIT_KHR | GL_SUBGROUP_FEATURE_VOTE_BIT_KHR |
+        GL_SUBGROUP_FEATURE_ARITHMETIC_BIT_KHR | GL_SUBGROUP_FEATURE_BALLOT_BIT_KHR |
+        GL_SUBGROUP_FEATURE_SHUFFLE_BIT_KHR | GL_SUBGROUP_FEATURE_SHUFFLE_RELATIVE_BIT_KHR |
+        GL_SUBGROUP_FEATURE_CLUSTERED_BIT_KHR | GL_SUBGROUP_FEATURE_QUAD_BIT_KHR;
+
+    inline Bool ShouldEmulateSubgroups(const Bool nativeSubgroupSupported) {
+        return MG_Config::Features.MagmaEmulateSubgroup && !nativeSubgroupSupported &&
+               !MG_Config::Features.DisableSubgroup;
+    }
+
+    inline Bool ShouldFixIterationRPSubgroupScratch() {
+        // Auto is ON: the patch is fingerprint-gated to iterationRP's reduction and
+        // grows one under-declared array; every other module passes through untouched.
+        return MG_Config::Features.FixIterationRPSubgroupScratch !=
+               MG_Config::QuirkOverride::ForceOff;
+    }
+
+    inline Bool ShouldFixIterationRPBarrier() {
+        return MG_Config::Features.IterationRPFixBarrier;
+    }
+
+    inline Bool ShouldDeriveNumSubgroups() {
+        // Auto is ON: gl_NumSubgroups must agree with the gl_SubgroupID range for the GL
+        // contract to hold, and the derived ceil() value is the one the renderer can pin
+        // with REQUIRE_FULL_SUBGROUPS - the driver builtin is the value with no
+        // cross-driver guarantee (Adreno returns 1 for an 8-subgroup dispatch).
+        return MG_Config::Features.DeriveNumSubgroups != MG_Config::QuirkOverride::ForceOff;
+    }
+} // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/VkIncludes.h

@@ -74,6 +74,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 // The context line (__VA_ARGS__ = its own format string + args) must be a SEPARATE log
 // call: appending its format to the base format while its arguments precede the base
 // arguments makes every conversion read the wrong slot (a %s pulling an int crashes).
+//
+// MGLOG_F and deliberately NOT latched. VK_VERIFY is the invariant-check macro: a Vulkan call
+// MobileGL believes it has already made legal came back non-success, which is a
+// should-never-happen state, not an expected failure mode a user hits. Those fast-fail loudly
+// and keep saying so - the log-quietness rules that latch W/E cover expected failures (driver
+// capability gaps, app misuse), not broken internal invariants. MOBILEGL_ASSERT below traps in
+// a DEBUG build; MGLOG_F is what makes the same condition visible in an INFO test run, where
+// the assert is compiled out by contract.
+//
+// A soft, recoverable failure must therefore NOT be routed through VK_VERIFY. Check the
+// VkResult directly and report it with MGLOG_E_ONCE - see VkTextureManager::SyncTextureResource,
+// where a driver legitimately refuses an image the format pre-check accepted.
 #define VK_VERIFY(expr, ...)                                                                                           \
     do {                                                                                                               \
         VkResult _vk_verify_result = (expr);                                                                           \

MobileGL/MG_Benchmark/CMakeLists.txt

@@ -42,3 +42,5 @@ set_tests_properties(SanityBench PROPERTIES LABELS benchmark)
 
 add_subdirectory(Program)
 add_subdirectory(Buffer)
+add_subdirectory(Driver)
+add_subdirectory(Container)

MobileGL/MG_Benchmark/Container/CMakeLists.txt

@@ -0,0 +1,20 @@
+cmake_minimum_required(VERSION 3.24)
+
+add_executable(
+        UnorderedMapBench
+        UnorderedMapBench.cpp
+)
+
+target_include_directories(UnorderedMapBench PRIVATE
+        ${MGL_ROOT}/include
+        ${MGL_ROOT}/MobileGL
+)
+
+target_link_libraries(
+        UnorderedMapBench PRIVATE
+        benchmark::benchmark
+        ${LINK_LIBRARIES}
+)
+
+add_test(NAME UnorderedMapBench COMMAND UnorderedMapBench --benchmark_counters_tabular=true)
+set_tests_properties(UnorderedMapBench PROPERTIES LABELS benchmark)

MobileGL/MG_Benchmark/Container/UnorderedMapBench.cpp

@@ -0,0 +1,248 @@
+// MobileGL - MobileGL/MG_Benchmark/Container/UnorderedMapBench.cpp
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+//
+// The standing performance observatory for MobileGL::UnorderedMap.
+//
+// This benchmarks the ALIAS, never a concrete table, so whatever UnorderedMap
+// names today is what gets measured - swap the container in MG_Util/Types.h and
+// re-run this same binary to get a directly comparable set of numbers. That is
+// the point of it: the container sits on per-draw paths, so a change to it needs
+// evidence, and the evidence should be produced the same way every time.
+//
+// The workloads are the shapes the tree actually exercises, not generic hash-map
+// microbenchmarks. Four key shapes, because they stress a hash function very
+// differently:
+//   * SEQUENTIAL   dense small integers - GL object names from the index generator
+//                  (buffer/texture/framebuffer/sampler registries).
+//   * POINTER      real heap addresses - StateBackendObjectRegistry keys on
+//                  StateObject*. These are aligned, so their low bits are the
+//                  least random part of the key; a table that indexes on raw low
+//                  bits clusters badly here and one that mixes first does not.
+//                  Taken from the real allocator rather than a synthetic stride,
+//                  which would flatter whichever table mixes its bits.
+//   * DIGEST       already well-mixed 64-bit values - the XXH64 pipeline,
+//                  vertex-input-state and program memos.
+//   * NAME         short strings - uniform/attribute name to location maps.
+//
+// Sizes sweep from 8 upward because the per-draw memos are usually SMALL; a table
+// that only wins at 4096 entries has not won anything that matters here.
+//
+// Run:  build-linux/MobileGL/MG_Benchmark/Container/UnorderedMapBench
+//   or: ctest -R UnorderedMapBench     (label: benchmark)
+
+#include <cstdint>
+#include <memory>
+#include <random>
+#include <string>
+#include <vector>
+#include <benchmark/benchmark.h>
+
+#include "MG_Util/Types.h"
+
+using namespace MobileGL;
+
+namespace {
+
+    constexpr Int64 kMinSize = 8;
+    constexpr Int64 kMaxSize = 4096;
+
+    // Keep the real allocations alive for the whole process: the POINTER shape is
+    // only honest if the keys are addresses the allocator actually handed out, and
+    // they have to stay unique (a freed address can be handed out twice).
+    std::vector<std::unique_ptr<char[]>>& PointerKeyStorage() {
+        static std::vector<std::unique_ptr<char[]>> storage;
+        return storage;
+    }
+
+    Vector<Uint64> SequentialKeys(SizeT n) {
+        Vector<Uint64> keys;
+        keys.reserve(n);
+        for (SizeT i = 0; i < n; ++i) keys.push_back(static_cast<Uint64>(i) + 1);
+        return keys;
+    }
+
+    Vector<Uint64> PointerKeys(SizeT n) {
+        auto& storage = PointerKeyStorage();
+        Vector<Uint64> keys;
+        keys.reserve(n);
+        std::mt19937_64 rng(0xBEEF);
+        std::vector<std::unique_ptr<char[]>> churn;
+        for (SizeT i = 0; i < n; ++i) {
+            // State objects are not all one size, and the allocator sees other
+            // traffic between them - a single uniform stride is not what this
+            // registry ever sees.
+            const SizeT sz = 96 + (rng() % 192);
+            auto p = std::make_unique<char[]>(sz);
+            keys.push_back(reinterpret_cast<Uint64>(p.get()));
+            storage.push_back(std::move(p));
+            if ((rng() & 3) == 0) churn.push_back(std::make_unique<char[]>(32 + (rng() % 128)));
+        }
+        return keys;
+    }
+
+    Vector<Uint64> DigestKeys(SizeT n) {
+        Vector<Uint64> keys;
+        keys.reserve(n);
+        std::mt19937_64 rng(0xC0FFEE);
+        for (SizeT i = 0; i < n; ++i) keys.push_back(rng());
+        return keys;
+    }
+
+    Vector<String> NameKeys(SizeT n) {
+        static const char* kPrefixes[] = {"u_", "a_", "mc_", "iris_", "gl_", "v_"};
+        Vector<String> keys;
+        keys.reserve(n);
+        for (SizeT i = 0; i < n; ++i) {
+            keys.push_back(String(kPrefixes[i % 6]) + "Uniform" + std::to_string(i) + "_xyz");
+        }
+        return keys;
+    }
+
+    // Key sets are built once per size and shared: generating them inside the timed
+    // loop would measure the generator (and, for POINTER, the allocator) instead of
+    // the table.
+    template <typename KeyVec, KeyVec (*Make)(SizeT)>
+    const KeyVec& CachedKeys(SizeT n) {
+        static UnorderedMap<SizeT, KeyVec> cache;
+        auto it = cache.find(n);
+        if (it != cache.end()) return it->second;
+        return cache.emplace(n, Make(n)).first->second;
+    }
+
+    template <typename Key>
+    UnorderedMap<Key, Uint64> Populated(const Vector<Key>& keys) {
+        UnorderedMap<Key, Uint64> map;
+        for (SizeT i = 0; i < keys.size(); ++i) map[keys[i]] = i;
+        return map;
+    }
+
+    // ---- the workloads ----------------------------------------------------
+
+    // The dominant per-draw operation by a wide margin: a populated cache that is
+    // read far more often than it is written.
+    template <typename KeyVec, KeyVec (*Make)(SizeT)>
+    void LookupHit(benchmark::State& state) {
+        const auto& keys = CachedKeys<KeyVec, Make>(static_cast<SizeT>(state.range(0)));
+        auto map = Populated(keys);
+        for (auto _ : state) {
+            for (const auto& k : keys) {
+                auto it = map.find(k);
+                benchmark::DoNotOptimize(it->second);
+            }
+        }
+        state.SetItemsProcessed(state.iterations() * static_cast<Int64>(keys.size()));
+    }
+
+    // "Is this resource cached yet?" answered NO - the probe length on a miss is a
+    // different cost from a hit, and resource caches ask this constantly.
+    template <typename KeyVec, KeyVec (*Make)(SizeT)>
+    void LookupMiss(benchmark::State& state) {
+        const SizeT n = static_cast<SizeT>(state.range(0));
+        const auto& keys = CachedKeys<KeyVec, Make>(n);
+        auto map = Populated(keys);
+        const KeyVec absent = Make(n); // same shape, never inserted
+        for (auto _ : state) {
+            for (const auto& k : absent) {
+                benchmark::DoNotOptimize(map.find(k) != map.end());
+            }
+        }
+        state.SetItemsProcessed(state.iterations() * static_cast<Int64>(absent.size()));
+    }
+
+    // Building a cache from empty, rehashes included.
+    template <typename KeyVec, KeyVec (*Make)(SizeT)>
+    void InsertGrow(benchmark::State& state) {
+        const auto& keys = CachedKeys<KeyVec, Make>(static_cast<SizeT>(state.range(0)));
+        for (auto _ : state) {
+            UnorderedMap<typename KeyVec::value_type, Uint64> map;
+            for (SizeT i = 0; i < keys.size(); ++i) map[keys[i]] = i;
+            benchmark::DoNotOptimize(map.size());
+        }
+        state.SetItemsProcessed(state.iterations() * static_cast<Int64>(keys.size()));
+    }
+
+    // Cache eviction and refill: erase half by key, put them back. This is the
+    // aged-out-entry sweep the pipeline and vertex-input caches do.
+    template <typename KeyVec, KeyVec (*Make)(SizeT)>
+    void EraseChurn(benchmark::State& state) {
+        const auto& keys = CachedKeys<KeyVec, Make>(static_cast<SizeT>(state.range(0)));
+        for (auto _ : state) {
+            state.PauseTiming();
+            auto map = Populated(keys);
+            state.ResumeTiming();
+            for (SizeT i = 0; i < keys.size(); i += 2) benchmark::DoNotOptimize(map.erase(keys[i]));
+            for (SizeT i = 0; i < keys.size(); i += 2) map[keys[i]] = i;
+            benchmark::DoNotOptimize(map.size());
+        }
+        state.SetItemsProcessed(state.iterations() * static_cast<Int64>(keys.size()));
+    }
+
+    // Mass eviction: erase-while-iterating across the whole table. This is the loop
+    // shape that a container's erase()-return contract can get wrong, and the one
+    // that fed garbage handles to vkDestroyPipeline when it was wrong before.
+    template <typename KeyVec, KeyVec (*Make)(SizeT)>
+    void EraseSweep(benchmark::State& state) {
+        const auto& keys = CachedKeys<KeyVec, Make>(static_cast<SizeT>(state.range(0)));
+        for (auto _ : state) {
+            state.PauseTiming();
+            auto map = Populated(keys);
+            state.ResumeTiming();
+            for (auto it = map.begin(); it != map.end();) it = map.erase(it);
+            benchmark::DoNotOptimize(map.size());
+        }
+        state.SetItemsProcessed(state.iterations() * static_cast<Int64>(keys.size()));
+    }
+
+    // Whole-table walks: the per-frame sweeps that age entries out, and the
+    // teardown loops that destroy every Vulkan object a cache owns.
+    template <typename KeyVec, KeyVec (*Make)(SizeT)>
+    void Iterate(benchmark::State& state) {
+        const auto& keys = CachedKeys<KeyVec, Make>(static_cast<SizeT>(state.range(0)));
+        auto map = Populated(keys);
+        for (auto _ : state) {
+            Uint64 acc = 0;
+            for (const auto& entry : map) acc += entry.second;
+            benchmark::DoNotOptimize(acc);
+        }
+        state.SetItemsProcessed(state.iterations() * static_cast<Int64>(keys.size()));
+    }
+
+} // namespace
+
+#define MGL_MAP_BENCH(WORKLOAD, SHAPE, VEC, MAKER)                                                 \
+    BENCHMARK_TEMPLATE(WORKLOAD, VEC, MAKER)                                                       \
+        ->Name(#WORKLOAD "/" #SHAPE)                                                               \
+        ->RangeMultiplier(8)                                                                       \
+        ->Range(kMinSize, kMaxSize)
+
+MGL_MAP_BENCH(LookupHit, sequential, Vector<Uint64>, SequentialKeys);
+MGL_MAP_BENCH(LookupHit, pointer, Vector<Uint64>, PointerKeys);
+MGL_MAP_BENCH(LookupHit, digest, Vector<Uint64>, DigestKeys);
+MGL_MAP_BENCH(LookupHit, name, Vector<String>, NameKeys);
+
+MGL_MAP_BENCH(LookupMiss, sequential, Vector<Uint64>, SequentialKeys);
+MGL_MAP_BENCH(LookupMiss, pointer, Vector<Uint64>, PointerKeys);
+MGL_MAP_BENCH(LookupMiss, digest, Vector<Uint64>, DigestKeys);
+MGL_MAP_BENCH(LookupMiss, name, Vector<String>, NameKeys);
+
+MGL_MAP_BENCH(InsertGrow, sequential, Vector<Uint64>, SequentialKeys);
+MGL_MAP_BENCH(InsertGrow, pointer, Vector<Uint64>, PointerKeys);
+MGL_MAP_BENCH(InsertGrow, digest, Vector<Uint64>, DigestKeys);
+MGL_MAP_BENCH(InsertGrow, name, Vector<String>, NameKeys);
+
+MGL_MAP_BENCH(EraseChurn, sequential, Vector<Uint64>, SequentialKeys);
+MGL_MAP_BENCH(EraseChurn, digest, Vector<Uint64>, DigestKeys);
+MGL_MAP_BENCH(EraseChurn, name, Vector<String>, NameKeys);
+
+MGL_MAP_BENCH(EraseSweep, sequential, Vector<Uint64>, SequentialKeys);
+MGL_MAP_BENCH(EraseSweep, digest, Vector<Uint64>, DigestKeys);
+
+MGL_MAP_BENCH(Iterate, sequential, Vector<Uint64>, SequentialKeys);
+MGL_MAP_BENCH(Iterate, digest, Vector<Uint64>, DigestKeys);
+
+BENCHMARK_MAIN();

MobileGL/MG_Benchmark/Driver/CMakeLists.txt

@@ -0,0 +1,15 @@
+cmake_minimum_required(VERSION 3.24)
+
+# A real, headless EGL client, deliberately NOT linked against MobileGL: it
+# dlopens one EGL provider at runtime ($DRIVERBENCH_EGL_LIB - the system
+# libEGL.so.1 for the native driver, or a libMobileGL.so path for either
+# MobileGL backend), so the same binary measures all three stacks.
+if (NOT UNIX OR APPLE OR ANDROID)
+    return()
+endif()
+
+add_executable(DriverBench DriverBench.c)
+target_link_libraries(DriverBench PRIVATE dl)
+
+add_test(NAME DriverBench COMMAND DriverBench draw_tiny)
+set_tests_properties(DriverBench PROPERTIES LABELS benchmark)

MobileGL/MG_Benchmark/Driver/DriverBench.c

@@ -0,0 +1,501 @@
+/* MobileGL - MobileGL/MG_Benchmark/Driver/DriverBench.c
+ * Copyright (c) 2025-2026 MobileGL-Dev
+ * Licensed under the GNU Lesser General Public License v3.0:
+ *   https://www.gnu.org/licenses/gpl-3.0.txt
+ *   https://www.gnu.org/licenses/lgpl-3.0.txt
+ * SPDX-License-Identifier: LGPL-3.0-only
+ * End of Source File Header
+ *
+ * Headless, EGL-based driver benchmark shaped like Minecraft's GL usage.
+ * Unlike the MobileGL_s microbenches next door this exercises a full GL
+ * stack: it dlopens ONE EGL provider ($DRIVERBENCH_EGL_LIB - the system
+ * libEGL.so.1 for the native driver, or a libMobileGL.so path for either
+ * MobileGL backend selected with MOBILEGL_BACKEND_TYPE), creates a desktop-GL
+ * context on a small pbuffer, renders into its own FBO and paces frames with
+ * glFinish. No window system is required: the default display is tried first
+ * so a desktop run reaches the real driver, and a headless box (CI, a build
+ * server) falls back to EGL_MESA_platform_surfaceless - see
+ * run_driver_bench.sh.
+ *
+ * Every case models one hot pattern from captured Minecraft traces:
+ *   draw_tiny        back-to-back glDrawElements, shared state (chunk batch)
+ *   draw_uniform     per-draw vec3 offset uniform + draw (chunk sections)
+ *   draw_multi_vao   per-draw VAO/VBO switch + draw (per-section buffers)
+ *   tex_pingpong     per-draw texture bind churn on one unit
+ *   program_pingpong alternate two programs + mat4 upload (chunk<->entity)
+ *   chunk_upload     glBufferData(NULL) orphan + glBufferSubData + draw
+ *   atlas_sprite     N 16x16 glTexSubImage2D into a 1024x512 atlas + draw
+ *   lightmap         full 16x16 lightmap respecify per frame + draw
+ *   scene_mix        composite frame built from the knobs below
+ *
+ * Output: one CSV line per case:
+ *   case,frames,ops_per_frame,median_frame_ms,ns_per_op,fps
+ */
+#include <dlfcn.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+
+/* ---- EGL constants ---- */
+typedef void* EGLDisplay;
+typedef void* EGLConfig;
+typedef void* EGLContext;
+typedef void* EGLSurface;
+typedef int EGLint;
+typedef unsigned int EGLBoolean;
+typedef unsigned int EGLenum;
+#define EGL_DEFAULT_DISPLAY ((void*)0)
+#define EGL_NO_CONTEXT ((EGLContext)0)
+#define EGL_NO_SURFACE ((EGLSurface)0)
+#define EGL_FALSE 0
+#define EGL_SURFACE_TYPE 0x3033
+#define EGL_PBUFFER_BIT 0x0001
+#define EGL_RENDERABLE_TYPE 0x3040
+#define EGL_OPENGL_BIT 0x0008
+#define EGL_RED_SIZE 0x3024
+#define EGL_GREEN_SIZE 0x3023
+#define EGL_BLUE_SIZE 0x3022
+#define EGL_DEPTH_SIZE 0x3025
+#define EGL_WIDTH 0x3057
+#define EGL_HEIGHT 0x3056
+#define EGL_NONE 0x3038
+#define EGL_OPENGL_API 0x30A2
+#define EGL_OPENGL_ES_API 0x30A0
+#define EGL_OPENGL_ES3_BIT 0x0040
+#define EGL_CONTEXT_CLIENT_VERSION 0x3098
+#define EGL_CONTEXT_MAJOR_VERSION 0x3098
+#define EGL_CONTEXT_MINOR_VERSION 0x30FB
+#define EGL_CONTEXT_OPENGL_PROFILE_MASK 0x30FD
+#define EGL_CONTEXT_OPENGL_CORE_PROFILE_BIT 0x00000001
+#define EGL_PLATFORM_SURFACELESS_MESA 0x31DD
+
+/* ---- GL constants ---- */
+#define GL_COLOR_BUFFER_BIT 0x00004000
+#define GL_DEPTH_BUFFER_BIT 0x00000100
+#define GL_TRIANGLES 0x0004
+#define GL_UNSIGNED_INT 0x1405
+#define GL_SHORT 0x1402
+#define GL_FLOAT 0x1406
+#define GL_UNSIGNED_BYTE 0x1401
+#define GL_ARRAY_BUFFER 0x8892
+#define GL_ELEMENT_ARRAY_BUFFER 0x8893
+#define GL_STATIC_DRAW 0x88E4
+#define GL_TEXTURE_2D 0x0DE1
+#define GL_TEXTURE0 0x84C0
+#define GL_RGBA 0x1908
+#define GL_RGBA8 0x8058
+#define GL_DEPTH_COMPONENT24 0x81A6
+#define GL_TEXTURE_MIN_FILTER 0x2801
+#define GL_TEXTURE_MAG_FILTER 0x2800
+#define GL_NEAREST 0x2600
+#define GL_NEAREST_MIPMAP_LINEAR 0x2702
+#define GL_DEPTH_TEST 0x0B71
+#define GL_BLEND 0x0BE2
+#define GL_SRC_ALPHA 0x0302
+#define GL_ONE_MINUS_SRC_ALPHA 0x0303
+#define GL_ONE 1
+#define GL_ZERO 0
+#define GL_VERTEX_SHADER 0x8B31
+#define GL_FRAGMENT_SHADER 0x8B30
+#define GL_COMPILE_STATUS 0x8B81
+#define GL_LINK_STATUS 0x8B82
+#define GL_VERSION 0x1F02
+#define GL_RENDERER 0x1F01
+#define GL_NO_ERROR 0
+#define GL_FRAMEBUFFER 0x8D40
+#define GL_RENDERBUFFER 0x8D41
+#define GL_COLOR_ATTACHMENT0 0x8CE0
+#define GL_DEPTH_ATTACHMENT 0x8D00
+#define GL_FRAMEBUFFER_COMPLETE 0x8CD5
+#define GL_SYNC_GPU_COMMANDS_COMPLETE 0x9117
+#define GL_SYNC_FLUSH_COMMANDS_BIT 0x00000001
+#define GL_UNIFORM_BUFFER 0x8A11
+#define GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT 0x8A34
+#define GL_DYNAMIC_DRAW 0x88E8
+#define GL_STREAM_DRAW 0x88E0
+#define GL_UNPACK_ALIGNMENT 0x0CF5
+#define GL_UNPACK_ROW_LENGTH 0x0CF2
+#define GL_UNPACK_SKIP_ROWS 0x0CF3
+#define GL_UNPACK_SKIP_PIXELS 0x0CF4
+#define GL_TEXTURE_WRAP_S 0x2802
+#define GL_TEXTURE_WRAP_T 0x2803
+#define GL_CLAMP_TO_EDGE 0x812F
+#define GL_REPEAT 0x2901
+
+typedef unsigned int GLuint;
+typedef int GLint;
+typedef int GLsizei;
+typedef unsigned int GLenum;
+typedef char GLchar;
+typedef unsigned char GLboolean;
+typedef long GLsizeiptr;
+typedef long GLintptr;
+
+/* ---- resolved entry points ---- */
+static void* (*g_eglGetProcAddress)(const char*);
+static void* g_provider;
+
+#define GLF(ret, name, args) static ret(*name) args;
+GLF(void, glClear, (unsigned))
+GLF(void, glClearColor, (float, float, float, float))
+GLF(void, glEnable, (GLenum))
+GLF(void, glDisable, (GLenum))
+GLF(void, glBlendFuncSeparate, (GLenum, GLenum, GLenum, GLenum))
+GLF(void, glDrawBuffers, (GLsizei, const GLenum*))
+GLF(void, glViewport, (GLint, GLint, GLsizei, GLsizei))
+GLF(const unsigned char*, glGetString, (GLenum))
+GLF(GLenum, glGetError, (void))
+GLF(void, glFinish, (void))
+GLF(void, glFlush, (void))
+GLF(void, glGenBuffers, (GLsizei, GLuint*))
+GLF(void, glBindBuffer, (GLenum, GLuint))
+GLF(void, glBufferData, (GLenum, GLsizeiptr, const void*, GLenum))
+GLF(void, glBufferSubData, (GLenum, GLintptr, GLsizeiptr, const void*))
+GLF(void, glGenVertexArrays, (GLsizei, GLuint*))
+GLF(void, glBindVertexArray, (GLuint))
+GLF(void, glEnableVertexAttribArray, (GLuint))
+GLF(void, glVertexAttribPointer, (GLuint, GLint, GLenum, GLboolean, GLsizei, const void*))
+GLF(void, glGenTextures, (GLsizei, GLuint*))
+GLF(void, glBindTexture, (GLenum, GLuint))
+GLF(void, glActiveTexture, (GLenum))
+GLF(void, glTexImage2D, (GLenum, GLint, GLint, GLsizei, GLsizei, GLint, GLenum, GLenum, const void*))
+GLF(void, glTexSubImage2D, (GLenum, GLint, GLint, GLint, GLsizei, GLsizei, GLenum, GLenum, const void*))
+GLF(void, glTexParameteri, (GLenum, GLenum, GLint))
+GLF(void, glPixelStorei, (GLenum, GLint))
+GLF(void, glGetIntegerv, (GLenum, GLint*))
+GLF(void, glGenerateMipmap, (GLenum))
+GLF(GLuint, glCreateShader, (GLenum))
+GLF(void, glShaderSource, (GLuint, GLsizei, const GLchar* const*, const GLint*))
+GLF(void, glCompileShader, (GLuint))
+GLF(void, glGetShaderiv, (GLuint, GLenum, GLint*))
+GLF(void, glGetShaderInfoLog, (GLuint, GLsizei, GLsizei*, GLchar*))
+GLF(GLuint, glCreateProgram, (void))
+GLF(void, glAttachShader, (GLuint, GLuint))
+GLF(void, glLinkProgram, (GLuint))
+GLF(void, glGetProgramiv, (GLuint, GLenum, GLint*))
+GLF(void, glUseProgram, (GLuint))
+GLF(GLint, glGetUniformLocation, (GLuint, const GLchar*))
+GLF(void, glUniform1i, (GLint, GLint))
+GLF(void, glUniform3f, (GLint, float, float, float))
+GLF(void, glUniformMatrix4fv, (GLint, GLsizei, GLboolean, const float*))
+GLF(void, glDrawElements, (GLenum, GLsizei, GLenum, const void*))
+GLF(void, glBindAttribLocation, (GLuint, GLuint, const GLchar*))
+GLF(void, glUniform3fv, (GLint, GLsizei, const float*))
+GLF(void, glDrawArrays, (GLenum, GLint, GLsizei))
+GLF(void, glDrawElementsBaseVertex, (GLenum, GLsizei, GLenum, const void*, GLint))
+GLF(void, glMultiDrawElementsBaseVertex,
+    (GLenum, const GLsizei*, GLenum, const void* const*, GLsizei, const GLint*))
+GLF(void, glBindBufferRange, (GLenum, GLuint, GLuint, GLintptr, GLsizeiptr))
+GLF(void, glBindBufferBase, (GLenum, GLuint, GLuint))
+GLF(GLuint, glGetUniformBlockIndex, (GLuint, const GLchar*))
+GLF(void, glUniformBlockBinding, (GLuint, GLuint, GLuint))
+GLF(void, glGenSamplers, (GLsizei, GLuint*))
+GLF(void, glBindSampler, (GLuint, GLuint))
+GLF(void, glSamplerParameteri, (GLuint, GLenum, GLint))
+GLF(void, glGenFramebuffers, (GLsizei, GLuint*))
+GLF(void, glBindFramebuffer, (GLenum, GLuint))
+GLF(void, glGenRenderbuffers, (GLsizei, GLuint*))
+GLF(void, glBindRenderbuffer, (GLenum, GLuint))
+GLF(void, glRenderbufferStorage, (GLenum, GLenum, GLsizei, GLsizei))
+GLF(void, glFramebufferRenderbuffer, (GLenum, GLenum, GLenum, GLuint))
+GLF(GLenum, glCheckFramebufferStatus, (GLenum))
+GLF(void*, glFenceSync, (GLenum, unsigned))
+GLF(GLenum, glClientWaitSync, (void*, unsigned, unsigned long long))
+GLF(void, glDeleteSync, (void*))
+
+static uint64_t now_ns(void) {
+    struct timespec ts;
+    clock_gettime(CLOCK_MONOTONIC, &ts);
+    return (uint64_t)ts.tv_sec * 1000000000ull + (uint64_t)ts.tv_nsec;
+}
+
+static int cmp_u64(const void* a, const void* b) {
+    uint64_t x = *(const uint64_t*)a, y = *(const uint64_t*)b;
+    return x < y ? -1 : x > y;
+}
+
+
+/* Scene, cases and the case table live next door so the Android plugin's
+ * in-process benchmark runs byte-identical bodies. */
+static void bench_gl_failed(const char* what, const char* detail) {
+    fprintf(stderr, "FAIL: %s %s\n", what, detail ? detail : "");
+    exit(1);
+}
+
+/* GLES has glDrawElementsBaseVertex (3.2 core) but no multi-draw form of it, so
+ * against a native mobile driver the multi-draw case issues the same sub-draws
+ * one at a time - which is what the extension folds up, and what an application
+ * without it would have to write. Desktop GL and MobileGL take the real call. */
+static void bench_multi_draw_elements_base_vertex(GLenum mode, const GLsizei* counts, GLenum type,
+                                                  const void* const* offsets, GLsizei drawCount,
+                                                  const GLint* baseVertices) {
+    if (glMultiDrawElementsBaseVertex) {
+        glMultiDrawElementsBaseVertex(mode, counts, type, offsets, drawCount, baseVertices);
+        return;
+    }
+    for (GLsizei i = 0; i < drawCount; ++i) {
+        glDrawElementsBaseVertex(mode, counts[i], type, offsets[i], baseVertices[i]);
+    }
+}
+
+#include "DriverBenchCases.inc"
+
+/* ---- bench driver: fence-paced frames on the offscreen FBO ----------------
+ * Frames are closed with a real fence wait, not glFinish: MobileGL implements
+ * glFinish and glFlush as no-ops (MG_Impl/GLImpl/Exporting/Definitions.cpp),
+ * so a glFinish-paced loop would time only the CPU-side submit on a MobileGL
+ * backend while timing submit-plus-GPU on the native driver - the two numbers
+ * would not describe the same work. A sync object is honoured by every stack
+ * measured here.
+ */
+typedef void (*case_fn)(int frame, long a, long b);
+static int g_warmup = 30, g_frames = 120;
+
+static void end_frame_wait(void) {
+    if (glFenceSync && glClientWaitSync && glDeleteSync) {
+        void* sync = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+        if (sync) {
+            glClientWaitSync(sync, GL_SYNC_FLUSH_COMMANDS_BIT, 1000000000ull);
+            glDeleteSync(sync);
+            return;
+        }
+    }
+    glFinish();
+}
+
+static void run_case(const char* name, case_fn body, long a, long b, long opsPerFrame) {
+    static uint64_t samples[4096];
+    if (g_frames > 4096) g_frames = 4096;
+    end_frame_wait();
+    for (int i = 0; i < g_warmup; ++i) {
+        glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+        body(i, a, b);
+        end_frame_wait();
+    }
+    for (int i = 0; i < g_frames; ++i) {
+        uint64_t t0 = now_ns();
+        glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+        body(i, a, b);
+        end_frame_wait();
+        samples[i] = now_ns() - t0;
+    }
+    qsort(samples, g_frames, sizeof(uint64_t), cmp_u64);
+    uint64_t med = samples[g_frames / 2];
+    double frameMs = med / 1e6;
+    double nsPerOp = opsPerFrame > 0 ? (double)med / (double)opsPerFrame : 0.0;
+    printf("%s,%d,%ld,%.3f,%.1f,%.1f\n", name, g_frames, opsPerFrame, frameMs, nsPerOp,
+           1e9 / (double)med);
+    fflush(stdout);
+    if (glGetError() != GL_NO_ERROR) fprintf(stderr, "WARN: GL error after %s\n", name);
+}
+
+/* A display that needs no window system. eglGetPlatformDisplay is EGL 1.5
+ * core and eglGetPlatformDisplayEXT is the EGL_EXT_platform_base spelling
+ * older loaders ship; both are client entry points, so they resolve before
+ * any display exists. Only the attribute-list types differ between the two
+ * and this passes none, so one cast covers both. */
+static EGLDisplay surfaceless_display(void) {
+    void* fn = dlsym(g_provider, "eglGetPlatformDisplay");
+    if (!fn) fn = g_eglGetProcAddress("eglGetPlatformDisplay");
+    if (!fn) fn = dlsym(g_provider, "eglGetPlatformDisplayEXT");
+    if (!fn) fn = g_eglGetProcAddress("eglGetPlatformDisplayEXT");
+    if (!fn) return NULL;
+    return ((EGLDisplay(*)(EGLenum, void*, const void*))fn)(EGL_PLATFORM_SURFACELESS_MESA,
+                                                            EGL_DEFAULT_DISPLAY, NULL);
+}
+
+/* ---- EGL bootstrap: one provider library, pbuffer, desktop-GL context ---- */
+static int boot_egl(void) {
+    const char* libpath = getenv("DRIVERBENCH_EGL_LIB");
+    if (!libpath) libpath = "libEGL.so.1";
+    g_provider = dlopen(libpath, RTLD_LAZY | RTLD_LOCAL);
+    if (!g_provider) {
+        fprintf(stderr, "FAIL: dlopen %s: %s\n", libpath, dlerror());
+        return 1;
+    }
+#define ESYM(name) \
+    void* p_##name = dlsym(g_provider, #name); \
+    if (!p_##name) { fprintf(stderr, "FAIL: dlsym %s\n", #name); return 1; }
+    ESYM(eglGetDisplay)
+    ESYM(eglInitialize)
+    ESYM(eglChooseConfig)
+    ESYM(eglBindAPI)
+    ESYM(eglCreateContext)
+    ESYM(eglCreatePbufferSurface)
+    ESYM(eglMakeCurrent)
+    ESYM(eglGetProcAddress)
+    ESYM(eglGetError)
+    g_eglGetProcAddress = (void* (*)(const char*))p_eglGetProcAddress;
+
+    EGLint (*getError)(void) = (EGLint(*)(void))p_eglGetError;
+    EGLBoolean (*initialize)(EGLDisplay, EGLint*, EGLint*) =
+        (EGLBoolean(*)(EGLDisplay, EGLint*, EGLint*))p_eglInitialize;
+
+    /* The default display first: it is the one a windowed app would get, and
+     * on a desktop it is the one that reaches the real GPU - which is the
+     * driver this bench exists to measure. It does need a window system,
+     * though; Mesa's default platform is X11, so with no $DISPLAY (CI, a
+     * build server, ssh without forwarding) eglInitialize fails. Fall back to
+     * EGL_MESA_platform_surfaceless rather than give up: every case draws into
+     * the FBO built by build_resources(), so no window is needed for any of
+     * the work being timed. */
+    EGLint maj = 0, min = 0;
+    const char* how = "default display";
+    EGLDisplay dpy = ((EGLDisplay(*)(void*))p_eglGetDisplay)(EGL_DEFAULT_DISPLAY);
+    if (!dpy || !initialize(dpy, &maj, &min)) {
+        dpy = surfaceless_display();
+        how = "surfaceless display";
+        if (!dpy || !initialize(dpy, &maj, &min)) {
+            fprintf(stderr, "FAIL: eglInitialize (0x%x)\n", getError());
+            return 1;
+        }
+    }
+    fprintf(stderr, "EGL %d.%d via %s (%s)\n", maj, min, libpath, how);
+
+    // Desktop GL first (that is what MobileGL exposes and what the cases are
+    // written against), GLES 3 second so the same binary can measure a device's
+    // native driver as the baseline. The .inc picks ESSL shader sources when the
+    // context turns out to be ES.
+    EGLBoolean (*chooseConfig)(EGLDisplay, const EGLint*, EGLConfig*, EGLint, EGLint*) =
+        (EGLBoolean(*)(EGLDisplay, const EGLint*, EGLConfig*, EGLint, EGLint*))p_eglChooseConfig;
+    EGLContext (*createContext)(EGLDisplay, EGLConfig, EGLContext, const EGLint*) =
+        (EGLContext(*)(EGLDisplay, EGLConfig, EGLContext, const EGLint*))p_eglCreateContext;
+    EGLBoolean (*bindApi)(EGLenum) = (EGLBoolean(*)(EGLenum))p_eglBindAPI;
+
+    EGLConfig cfg = NULL;
+    EGLint ncfg = 0;
+    EGLContext ctx = EGL_NO_CONTEXT;
+
+    if (bindApi(EGL_OPENGL_API)) {
+        const EGLint cfgAttribs[] = {EGL_SURFACE_TYPE, EGL_PBUFFER_BIT, EGL_RED_SIZE, 8,
+                                     EGL_DEPTH_SIZE, 24, EGL_RENDERABLE_TYPE, EGL_OPENGL_BIT, EGL_NONE};
+        if (chooseConfig(dpy, cfgAttribs, &cfg, 1, &ncfg) && ncfg >= 1) {
+            const EGLint ctxAttribs[] = {EGL_CONTEXT_MAJOR_VERSION, 3, EGL_CONTEXT_MINOR_VERSION, 2,
+                                         EGL_CONTEXT_OPENGL_PROFILE_MASK,
+                                         EGL_CONTEXT_OPENGL_CORE_PROFILE_BIT, EGL_NONE};
+            ctx = createContext(dpy, cfg, EGL_NO_CONTEXT, ctxAttribs);
+            if (ctx == EGL_NO_CONTEXT) ctx = createContext(dpy, cfg, EGL_NO_CONTEXT, NULL);
+        }
+    }
+    if (ctx == EGL_NO_CONTEXT) {
+        if (!bindApi(EGL_OPENGL_ES_API)) {
+            fprintf(stderr, "FAIL: neither OpenGL nor OpenGL ES is bindable on this provider\n");
+            return 1;
+        }
+        const EGLint esCfgAttribs[] = {EGL_SURFACE_TYPE, EGL_PBUFFER_BIT, EGL_RED_SIZE, 8,
+                                       EGL_GREEN_SIZE, 8, EGL_BLUE_SIZE, 8, EGL_DEPTH_SIZE, 24,
+                                       EGL_RENDERABLE_TYPE, EGL_OPENGL_ES3_BIT, EGL_NONE};
+        ncfg = 0;
+        if (!chooseConfig(dpy, esCfgAttribs, &cfg, 1, &ncfg) || ncfg < 1) {
+            // EGL_SURFACE_TYPE 0 matches any config: a stack that offers no
+            // pbuffer at all is still usable through the surfaceless context
+            // path below.
+            const EGLint relaxed[] = {EGL_SURFACE_TYPE, 0, EGL_RED_SIZE, 8, EGL_NONE};
+            if (!chooseConfig(dpy, relaxed, &cfg, 1, &ncfg) || ncfg < 1) {
+                fprintf(stderr, "FAIL: eglChooseConfig\n");
+                return 1;
+            }
+        }
+        const EGLint esCtxAttribs[] = {EGL_CONTEXT_CLIENT_VERSION, 3, EGL_NONE};
+        ctx = createContext(dpy, cfg, EGL_NO_CONTEXT, esCtxAttribs);
+    }
+    if (ctx == EGL_NO_CONTEXT) {
+        fprintf(stderr, "FAIL: eglCreateContext (0x%x)\n", getError());
+        return 1;
+    }
+
+    /* The pbuffer only exists to have something to make current - nothing is
+     * ever drawn to it. Where there is no pbuffer config, EGL_NO_SURFACE is
+     * exactly what EGL_KHR_surfaceless_context takes, so the same call covers
+     * both. */
+    const EGLint pbAttribs[] = {EGL_WIDTH, 64, EGL_HEIGHT, 64, EGL_NONE};
+    EGLSurface surf = ((EGLSurface(*)(EGLDisplay, EGLConfig, const EGLint*))p_eglCreatePbufferSurface)(
+        dpy, cfg, pbAttribs);
+    if (surf == EGL_NO_SURFACE)
+        fprintf(stderr, "no pbuffer (0x%x), using a surfaceless context\n", getError());
+    if (!((EGLBoolean(*)(EGLDisplay, EGLSurface, EGLSurface, EGLContext))p_eglMakeCurrent)(dpy, surf,
+                                                                                          surf, ctx)) {
+        fprintf(stderr, "FAIL: eglMakeCurrent (0x%x)\n", getError());
+        return 1;
+    }
+
+    /* Core GL entry points: eglGetProcAddress first (EGL 1.5 serves core
+     * functions), provider dlsym as fallback (both glvnd and MobileGL export
+     * the gl* symbols directly). */
+#define RESOLVE(name)                                                              \
+    do {                                                                           \
+        *(void**)&name = g_eglGetProcAddress(#name);                               \
+        if (!name) *(void**)&name = dlsym(g_provider, #name);                      \
+        if (!name) { fprintf(stderr, "FAIL: resolve %s\n", #name); return 1; }     \
+    } while (0)
+    RESOLVE(glClear); RESOLVE(glClearColor); RESOLVE(glEnable); RESOLVE(glViewport);
+    RESOLVE(glDisable); RESOLVE(glBlendFuncSeparate); RESOLVE(glDrawBuffers);
+    RESOLVE(glGetString); RESOLVE(glGetError); RESOLVE(glFinish); RESOLVE(glFlush);
+    RESOLVE(glGenBuffers); RESOLVE(glBindBuffer); RESOLVE(glBufferData); RESOLVE(glBufferSubData);
+    RESOLVE(glGenVertexArrays); RESOLVE(glBindVertexArray); RESOLVE(glEnableVertexAttribArray);
+    RESOLVE(glVertexAttribPointer); RESOLVE(glGenTextures); RESOLVE(glBindTexture);
+    RESOLVE(glActiveTexture); RESOLVE(glTexImage2D); RESOLVE(glTexSubImage2D);
+    RESOLVE(glTexParameteri); RESOLVE(glGenerateMipmap); RESOLVE(glCreateShader);
+    RESOLVE(glPixelStorei); RESOLVE(glGetIntegerv);
+    RESOLVE(glShaderSource); RESOLVE(glCompileShader); RESOLVE(glGetShaderiv);
+    RESOLVE(glGetShaderInfoLog); RESOLVE(glCreateProgram); RESOLVE(glAttachShader);
+    RESOLVE(glLinkProgram); RESOLVE(glGetProgramiv); RESOLVE(glUseProgram);
+    RESOLVE(glGetUniformLocation); RESOLVE(glUniform1i); RESOLVE(glUniform3f);
+    RESOLVE(glUniformMatrix4fv); RESOLVE(glDrawElements); RESOLVE(glBindAttribLocation);
+    RESOLVE(glUniform3fv); RESOLVE(glDrawArrays); RESOLVE(glDrawElementsBaseVertex);
+    RESOLVE(glBindBufferRange); RESOLVE(glBindBufferBase);
+    RESOLVE(glGetUniformBlockIndex); RESOLVE(glUniformBlockBinding);
+    RESOLVE(glGenSamplers); RESOLVE(glBindSampler); RESOLVE(glSamplerParameteri);
+    RESOLVE(glGenFramebuffers); RESOLVE(glBindFramebuffer); RESOLVE(glGenRenderbuffers);
+    RESOLVE(glBindRenderbuffer); RESOLVE(glRenderbufferStorage); RESOLVE(glFramebufferRenderbuffer);
+    RESOLVE(glCheckFramebufferStatus);
+    // Optional: end_frame_wait() falls back to glFinish when a stack has no
+    // sync objects, so resolve without failing the run.
+    *(void**)&glFenceSync = g_eglGetProcAddress("glFenceSync");
+    if (!glFenceSync) *(void**)&glFenceSync = dlsym(g_provider, "glFenceSync");
+    *(void**)&glClientWaitSync = g_eglGetProcAddress("glClientWaitSync");
+    if (!glClientWaitSync) *(void**)&glClientWaitSync = dlsym(g_provider, "glClientWaitSync");
+    *(void**)&glDeleteSync = g_eglGetProcAddress("glDeleteSync");
+    if (!glDeleteSync) *(void**)&glDeleteSync = dlsym(g_provider, "glDeleteSync");
+    // Desktop-only: GLES 3.2 has DrawElementsBaseVertex but no multi-draw form,
+    // so bench_multi_draw_elements_base_vertex() emulates it when this is null.
+    *(void**)&glMultiDrawElementsBaseVertex = g_eglGetProcAddress("glMultiDrawElementsBaseVertex");
+    if (!glMultiDrawElementsBaseVertex)
+        *(void**)&glMultiDrawElementsBaseVertex = dlsym(g_provider, "glMultiDrawElementsBaseVertex");
+
+    fprintf(stderr, "renderer: %s\n", glGetString(GL_RENDERER));
+    fprintf(stderr, "version:  %s\n", glGetString(GL_VERSION));
+    return 0;
+}
+
+int main(int argc, char** argv) {
+    long draws = 2048;
+    if (getenv("DRIVERBENCH_DRAWS")) draws = atol(getenv("DRIVERBENCH_DRAWS"));
+    if (getenv("DRIVERBENCH_FRAMES")) g_frames = atoi(getenv("DRIVERBENCH_FRAMES"));
+    if (getenv("DRIVERBENCH_SPRITES")) g_mixSprites = atol(getenv("DRIVERBENCH_SPRITES"));
+
+    if (boot_egl()) return 1;
+    build_resources();
+
+    printf("case,frames,ops_per_frame,median_frame_ms,ns_per_op,fps\n");
+    for (int i = 0; i < kBenchCaseCount; ++i) {
+        const BenchCaseDesc* c = &kBenchCases[i];
+        if (argc > 1) {
+            int wanted = 0;
+            for (int j = 1; j < argc; ++j)
+                if (strcmp(argv[j], c->name) == 0) wanted = 1;
+            if (!wanted) continue;
+        }
+        // The generic cases scale with DRIVERBENCH_DRAWS; the mc_* rates are
+        // measured and must not move, or the numbers stop being comparable.
+        long a = c->a, ops = c->opsPerFrame;
+        if (strncmp(c->name, "mc_", 3) != 0 && a > 100) {
+            a = draws * a / 2048;
+            ops = c->opsPerFrame * draws / 2048;
+        }
+        run_case(c->name, c->fn, a, c->b, ops);
+    }
+    return 0;
+}

MobileGL/MG_Benchmark/Driver/DriverBenchCases.inc

@@ -0,0 +1,640 @@
+/* MobileGL - MobileGL/MG_Benchmark/Driver/DriverBenchCases.inc
+ * Copyright (c) 2025-2026 MobileGL-Dev
+ * Licensed under the GNU Lesser General Public License v3.0:
+ *   https://www.gnu.org/licenses/gpl-3.0.txt
+ *   https://www.gnu.org/licenses/lgpl-3.0.txt
+ * SPDX-License-Identifier: LGPL-3.0-only
+ * End of Source File Header
+ *
+ * The benchmark scene and its cases, with no harness and no GL loader: the
+ * includer supplies both. DriverBench.c drives it through function pointers
+ * resolved from one EGL provider; MG_Util/SelfTest/DriverBenchJni.cpp drives
+ * it through MobileGL's own frontend entry points inside the Android plugin.
+ * Sharing the bodies is the point - a number from the phone and a number from
+ * the desktop have to describe the same work.
+ *
+ * The includer must have declared, before including this file: the GL types
+ * and enums used below, and callable gl* entry points with the standard
+ * signatures. bench_gl_failed() is called (and must be defined) when shader
+ * compilation or linking fails, so a caller can report the failure instead of
+ * dying inside a benchmark.
+ */
+
+/* ---- shared scene resources (Minecraft-shaped) ---- */
+#define MAX_SECTIONS 512
+static GLuint g_progChunk, g_progEntity;
+static GLint g_uOffsetChunk, g_uMvpChunk, g_uMvpEntity;
+static GLuint g_vao[MAX_SECTIONS], g_vbo[MAX_SECTIONS];
+static GLuint g_sharedIbo;
+static GLuint g_texAtlas, g_texLight, g_texEntity;
+static int g_quadsPerSection = 128; /* 128 quads = 512 verts, 768 indices */
+static unsigned char* g_scratch;
+/* Uniform ring + sampler for the 26.2-shaped cases (see the case block below). */
+static GLuint g_uboRing;
+static GLint g_uboAlign = 256;
+static size_t g_uboSlot = 256;
+static GLuint g_sampler;
+/* Two small offscreen targets for the 26.2-style render-pass churn case. */
+static GLuint g_passFbo[2];
+static GLuint g_passColor[2];
+static float g_mvp[16] = {0.002f, 0, 0, 0, 0, 0.002f, 0, 0, 0, 0, -0.001f, 0, -1.f, -1.f, 0.f, 1.f};
+
+/* Minecraft chunk vertex: pos 3f, color 4ub, uv 2f, packed light 2s -> 32 B */
+#define VERT_STRIDE 32
+static void fill_section_vertices(unsigned char* dst, int quads, unsigned seed) {
+    for (int q = 0; q < quads * 4; ++q) {
+        float* f = (float*)(dst + q * VERT_STRIDE);
+        unsigned r = seed = seed * 1664525u + 1013904223u;
+        f[0] = (float)(q & 31) * 8.0f + (float)(r & 7);
+        f[1] = (float)((q >> 5) & 31) * 8.0f;
+        f[2] = (float)(q % 7) * 0.1f;
+        dst[q * VERT_STRIDE + 12] = (unsigned char)r;
+        dst[q * VERT_STRIDE + 13] = (unsigned char)(r >> 8);
+        dst[q * VERT_STRIDE + 14] = (unsigned char)(r >> 16);
+        dst[q * VERT_STRIDE + 15] = 255;
+        f[4] = (float)(r & 1023) / 1024.0f;
+        f[5] = (float)((r >> 10) & 511) / 512.0f;
+        ((short*)(dst + q * VERT_STRIDE + 24))[0] = 15 << 4;
+        ((short*)(dst + q * VERT_STRIDE + 24))[1] = 15 << 4;
+    }
+}
+
+static GLuint make_shader(GLenum kind, const char* src) {
+    GLuint sh = glCreateShader(kind);
+    glShaderSource(sh, 1, &src, NULL);
+    glCompileShader(sh);
+    GLint ok = 0;
+    glGetShaderiv(sh, GL_COMPILE_STATUS, &ok);
+    if (!ok) {
+        char log[1024];
+        glGetShaderInfoLog(sh, sizeof log, NULL, log);
+        bench_gl_failed("shader compile", log);
+        return 0;
+    }
+    return sh;
+}
+
+static GLuint make_program(const char* vs_src, const char* fs_src) {
+    GLuint prog = glCreateProgram();
+    glAttachShader(prog, make_shader(GL_VERTEX_SHADER, vs_src));
+    glAttachShader(prog, make_shader(GL_FRAGMENT_SHADER, fs_src));
+    glBindAttribLocation(prog, 0, "aPos");
+    glBindAttribLocation(prog, 1, "aColor");
+    glBindAttribLocation(prog, 2, "aUv");
+    glBindAttribLocation(prog, 3, "aLight");
+    glLinkProgram(prog);
+    GLint ok = 0;
+    glGetProgramiv(prog, GL_LINK_STATUS, &ok);
+    if (!ok) {
+        bench_gl_failed("program link", "");
+        return 0;
+    }
+    return prog;
+}
+
+static const char* kChunkVs =
+    "#version 150 core\n"
+    "in vec3 aPos; in vec4 aColor; in vec2 aUv; in vec2 aLight;\n"
+    "uniform mat4 uMvp; uniform vec3 uOffset;\n"
+    "out vec4 vColor; out vec2 vUv; out vec2 vLight;\n"
+    "void main(){ gl_Position = uMvp * vec4(aPos + uOffset, 1.0);\n"
+    "  vColor = aColor; vUv = aUv; vLight = aLight * (1.0/256.0); }\n";
+static const char* kChunkFs =
+    "#version 150 core\n"
+    "in vec4 vColor; in vec2 vUv; in vec2 vLight; out vec4 o;\n"
+    "uniform sampler2D uAtlas; uniform sampler2D uLight;\n"
+    "void main(){ o = texture(uAtlas, vUv) * vColor * texture(uLight, vLight); }\n";
+static const char* kEntityVs =
+    "#version 150 core\n"
+    "in vec3 aPos; in vec4 aColor; in vec2 aUv; in vec2 aLight;\n"
+    "uniform mat4 uMvp; uniform mat4 uModel;\n"
+    "out vec4 vColor; out vec2 vUv;\n"
+    "void main(){ gl_Position = uMvp * uModel * vec4(aPos, 1.0); vColor = aColor; vUv = aUv; }\n";
+static const char* kEntityFs =
+    "#version 150 core\n"
+    "in vec4 vColor; in vec2 vUv; out vec4 o; uniform sampler2D uTex;\n"
+    "void main(){ o = texture(uTex, vUv) * vColor; }\n";
+
+// ESSL 3.20 twins of the four shaders above. The bodies are identical; only the
+// version line and the precision qualifiers differ, so the two paths compile the
+// same work. Needed because this bench also runs against a device's native GLES
+// driver as the baseline MobileGL is measured against, and that driver rejects
+// desktop GLSL - while MobileGL is fed desktop GLSL on purpose, since translating
+// it is the thing under test.
+static const char* kChunkVsEs =
+    "#version 320 es\n"
+    "precision highp float;\n"
+    "in vec3 aPos; in vec4 aColor; in vec2 aUv; in vec2 aLight;\n"
+    "uniform mat4 uMvp; uniform vec3 uOffset;\n"
+    "out vec4 vColor; out vec2 vUv; out vec2 vLight;\n"
+    "void main(){ gl_Position = uMvp * vec4(aPos + uOffset, 1.0);\n"
+    "  vColor = aColor; vUv = aUv; vLight = aLight * (1.0/256.0); }\n";
+static const char* kChunkFsEs =
+    "#version 320 es\n"
+    "precision mediump float;\n"
+    "in vec4 vColor; in vec2 vUv; in vec2 vLight; out vec4 o;\n"
+    "uniform sampler2D uAtlas; uniform sampler2D uLight;\n"
+    "void main(){ o = texture(uAtlas, vUv) * vColor * texture(uLight, vLight); }\n";
+static const char* kEntityVsEs =
+    "#version 320 es\n"
+    "precision highp float;\n"
+    "in vec3 aPos; in vec4 aColor; in vec2 aUv; in vec2 aLight;\n"
+    "uniform mat4 uMvp; uniform mat4 uModel;\n"
+    "out vec4 vColor; out vec2 vUv;\n"
+    "void main(){ gl_Position = uMvp * uModel * vec4(aPos, 1.0); vColor = aColor; vUv = aUv; }\n";
+static const char* kEntityFsEs =
+    "#version 320 es\n"
+    "precision mediump float;\n"
+    "in vec4 vColor; in vec2 vUv; out vec4 o; uniform sampler2D uTex;\n"
+    "void main(){ o = texture(uTex, vUv) * vColor; }\n";
+
+// True once build_resources() has seen a GL_VERSION beginning with "OpenGL ES".
+static int g_isGlesContext = 0;
+
+static void setup_vao(GLuint vao, GLuint vbo, GLuint ibo) {
+    glBindVertexArray(vao);
+    glBindBuffer(GL_ARRAY_BUFFER, vbo);
+    glEnableVertexAttribArray(0);
+    glEnableVertexAttribArray(1);
+    glEnableVertexAttribArray(2);
+    glEnableVertexAttribArray(3);
+    glVertexAttribPointer(0, 3, GL_FLOAT, 0, VERT_STRIDE, (void*)0);
+    glVertexAttribPointer(1, 4, GL_UNSIGNED_BYTE, 1, VERT_STRIDE, (void*)12);
+    glVertexAttribPointer(2, 2, GL_FLOAT, 0, VERT_STRIDE, (void*)16);
+    glVertexAttribPointer(3, 2, GL_SHORT, 0, VERT_STRIDE, (void*)24);
+    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, ibo);
+}
+
+static GLuint g_mainFbo;
+
+static void build_resources(void) {
+    /* offscreen render target: 1280x720 RBO FBO, like CTS fbo surface mode */
+    GLuint fbo, rboColor, rboDepth;
+    glGenFramebuffers(1, &fbo);
+    g_mainFbo = fbo;
+    glBindFramebuffer(GL_FRAMEBUFFER, fbo);
+    glGenRenderbuffers(1, &rboColor);
+    glBindRenderbuffer(GL_RENDERBUFFER, rboColor);
+    glRenderbufferStorage(GL_RENDERBUFFER, GL_RGBA8, 1280, 720);
+    glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, rboColor);
+    glGenRenderbuffers(1, &rboDepth);
+    glBindRenderbuffer(GL_RENDERBUFFER, rboDepth);
+    glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH_COMPONENT24, 1280, 720);
+    glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, rboDepth);
+    if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+        bench_gl_failed("FBO incomplete", "");
+        return;
+    }
+
+    const char* versionString = (const char*)glGetString(GL_VERSION);
+    g_isGlesContext = versionString != NULL && strncmp(versionString, "OpenGL ES", 9) == 0;
+    g_progChunk = g_isGlesContext ? make_program(kChunkVsEs, kChunkFsEs) : make_program(kChunkVs, kChunkFs);
+    g_progEntity = g_isGlesContext ? make_program(kEntityVsEs, kEntityFsEs) : make_program(kEntityVs, kEntityFs);
+    glUseProgram(g_progChunk);
+    g_uMvpChunk = glGetUniformLocation(g_progChunk, "uMvp");
+    g_uOffsetChunk = glGetUniformLocation(g_progChunk, "uOffset");
+    glUniform1i(glGetUniformLocation(g_progChunk, "uAtlas"), 0);
+    glUniform1i(glGetUniformLocation(g_progChunk, "uLight"), 2);
+    glUniformMatrix4fv(g_uMvpChunk, 1, 0, g_mvp);
+    glUseProgram(g_progEntity);
+    g_uMvpEntity = glGetUniformLocation(g_progEntity, "uMvp");
+    glUniform1i(glGetUniformLocation(g_progEntity, "uTex"), 0);
+    glUniformMatrix4fv(g_uMvpEntity, 1, 0, g_mvp);
+    glUseProgram(g_progChunk);
+
+    /* shared quad index buffer, like Blaze3D's RenderSystem shared sequences */
+    int maxQuads = 4096;
+    unsigned* idx = (unsigned*)malloc((size_t)maxQuads * 6 * 4);
+    for (int q = 0; q < maxQuads; ++q) {
+        unsigned base = q * 4;
+        unsigned* p = idx + q * 6;
+        p[0] = base; p[1] = base + 1; p[2] = base + 2;
+        p[3] = base + 2; p[4] = base + 3; p[5] = base;
+    }
+    glGenBuffers(1, &g_sharedIbo);
+    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, g_sharedIbo);
+    glBufferData(GL_ELEMENT_ARRAY_BUFFER, maxQuads * 6 * 4, idx, GL_STATIC_DRAW);
+    free(idx);
+
+    g_scratch = (unsigned char*)malloc(4 * 1024 * 1024);
+    memset(g_scratch, 0x5a, 4 * 1024 * 1024);
+
+    glGenVertexArrays(MAX_SECTIONS, g_vao);
+    glGenBuffers(MAX_SECTIONS, g_vbo);
+    int bytes = g_quadsPerSection * 4 * VERT_STRIDE;
+    for (int i = 0; i < MAX_SECTIONS; ++i) {
+        fill_section_vertices(g_scratch, g_quadsPerSection, i * 7919u + 1);
+        glBindBuffer(GL_ARRAY_BUFFER, g_vbo[i]);
+        glBufferData(GL_ARRAY_BUFFER, bytes, g_scratch, GL_STATIC_DRAW);
+        setup_vao(g_vao[i], g_vbo[i], g_sharedIbo);
+    }
+
+    glGenTextures(1, &g_texAtlas);
+    glActiveTexture(GL_TEXTURE0);
+    glBindTexture(GL_TEXTURE_2D, g_texAtlas);
+    glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, 1024, 512, 0, GL_RGBA, GL_UNSIGNED_BYTE, g_scratch);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST_MIPMAP_LINEAR);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+    glGenerateMipmap(GL_TEXTURE_2D);
+
+    glGenTextures(1, &g_texLight);
+    glActiveTexture(GL_TEXTURE0 + 2);
+    glBindTexture(GL_TEXTURE_2D, g_texLight);
+    glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, 16, 16, 0, GL_RGBA, GL_UNSIGNED_BYTE, g_scratch);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+
+    glGenTextures(1, &g_texEntity);
+    glActiveTexture(GL_TEXTURE0);
+    glBindTexture(GL_TEXTURE_2D, g_texEntity);
+    glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, 64, 64, 0, GL_RGBA, GL_UNSIGNED_BYTE, g_scratch);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+    glBindTexture(GL_TEXTURE_2D, g_texAtlas);
+
+    // Uniform ring the 26.2-style case sub-ranges into, sized like a real
+    // frame's worth of per-draw uniform slots.
+    GLint align = 256;
+    glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &align);
+    g_uboAlign = align > 0 ? align : 256;
+    g_uboSlot = (size_t)g_uboAlign;
+    glGenBuffers(1, &g_uboRing);
+    glBindBuffer(GL_UNIFORM_BUFFER, g_uboRing);
+    glBufferData(GL_UNIFORM_BUFFER, 4 * 1024 * 1024, g_scratch, GL_DYNAMIC_DRAW);
+    glBindBuffer(GL_UNIFORM_BUFFER, 0);
+
+    for (int i = 0; i < 2; ++i) {
+        glGenFramebuffers(1, &g_passFbo[i]);
+        glBindFramebuffer(GL_FRAMEBUFFER, g_passFbo[i]);
+        glGenRenderbuffers(1, &g_passColor[i]);
+        glBindRenderbuffer(GL_RENDERBUFFER, g_passColor[i]);
+        glRenderbufferStorage(GL_RENDERBUFFER, GL_RGBA8, 256, 256);
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, g_passColor[i]);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            bench_gl_failed("pass FBO incomplete", "");
+            return;
+        }
+    }
+    /* back to the main offscreen target the harness set up */
+    glBindFramebuffer(GL_FRAMEBUFFER, g_mainFbo);
+
+    glGenSamplers(1, &g_sampler);
+    glSamplerParameteri(g_sampler, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+    glSamplerParameteri(g_sampler, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+
+    glEnable(GL_DEPTH_TEST);
+    glClearColor(0.3f, 0.5f, 0.9f, 1.0f);
+    glViewport(0, 0, 1280, 720);
+    const GLenum setupError = glGetError();
+    if (setupError != GL_NO_ERROR) {
+        char message[64];
+        snprintf(message, sizeof message, "0x%04x", setupError);
+        bench_gl_failed("GL error during resource setup", message);
+    }
+}
+
+static void case_draw_tiny(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+}
+
+static void case_draw_uniform(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) {
+        glUniform3f(g_uOffsetChunk, (float)(i & 15), (float)((i >> 4) & 15), 0.0f);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+}
+
+static void case_draw_multi_vao(int frame, long a, long b) {
+    (void)frame; (void)b;
+    for (long i = 0; i < a; ++i) {
+        glBindVertexArray(g_vao[i % MAX_SECTIONS]);
+        glUniform3f(g_uOffsetChunk, (float)(i & 15), (float)((i >> 4) & 15), 0.0f);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+}
+
+static void case_tex_pingpong(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) {
+        glBindTexture(GL_TEXTURE_2D, (i & 1) ? g_texEntity : g_texAtlas);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+    glBindTexture(GL_TEXTURE_2D, g_texAtlas);
+}
+
+static void case_program_pingpong(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) {
+        if (i & 1) {
+            glUseProgram(g_progEntity);
+            glUniformMatrix4fv(g_uMvpEntity, 1, 0, g_mvp);
+        } else {
+            glUseProgram(g_progChunk);
+            glUniform3f(g_uOffsetChunk, (float)(i & 15), 0.0f, 0.0f);
+        }
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+    glUseProgram(g_progChunk);
+}
+
+/* a = uploads per frame, b = bytes per upload (0 => section size) */
+static void case_chunk_upload(int frame, long a, long b) {
+    if (b <= 0) b = g_quadsPerSection * 4 * VERT_STRIDE;
+    if (b > 4 * 1024 * 1024) b = 4 * 1024 * 1024;
+    for (long i = 0; i < a; ++i) {
+        int slot = (int)(((long)frame * a + i) % MAX_SECTIONS);
+        glBindBuffer(GL_ARRAY_BUFFER, g_vbo[slot]);
+        glBufferData(GL_ARRAY_BUFFER, b, NULL, GL_STATIC_DRAW); /* orphan */
+        glBufferSubData(GL_ARRAY_BUFFER, 0, b, g_scratch);
+        glBindVertexArray(g_vao[slot]);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+}
+
+/* a = sprite updates per frame */
+static void case_atlas_sprite(int frame, long a, long b) {
+    (void)b;
+    glBindVertexArray(g_vao[0]);
+    glBindTexture(GL_TEXTURE_2D, g_texAtlas);
+    for (long i = 0; i < a; ++i) {
+        int x = (int)((frame * 13 + i * 17) % (1024 - 16));
+        int y = (int)((frame * 7 + i * 29) % (512 - 16));
+        glTexSubImage2D(GL_TEXTURE_2D, 0, x, y, 16, 16, GL_RGBA, GL_UNSIGNED_BYTE, g_scratch);
+    }
+    glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+}
+
+/* a = lightmap updates (+draw) per frame */
+static void case_lightmap(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) {
+        glActiveTexture(GL_TEXTURE0 + 2);
+        glBindTexture(GL_TEXTURE_2D, g_texLight);
+        glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, 16, 16, GL_RGBA, GL_UNSIGNED_BYTE, g_scratch);
+        glActiveTexture(GL_TEXTURE0);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+}
+
+/* Composite: a = total draws, b = uploads per frame. Mix modeled on trace
+ * analysis: chunk draws with per-draw offset uniform across sections, 10%
+ * entity-style program flips, per-frame lightmap + sprite updates, b chunk
+ * re-uploads. */
+static long g_mixSprites = 8;
+static void case_scene_mix(int frame, long a, long b) {
+    glActiveTexture(GL_TEXTURE0 + 2);
+    glBindTexture(GL_TEXTURE_2D, g_texLight);
+    glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, 16, 16, GL_RGBA, GL_UNSIGNED_BYTE, g_scratch);
+    glActiveTexture(GL_TEXTURE0);
+    glBindTexture(GL_TEXTURE_2D, g_texAtlas);
+    for (long i = 0; i < g_mixSprites; ++i) {
+        int x = (int)((frame * 13 + i * 17) % (1024 - 16));
+        int y = (int)((frame * 7 + i * 29) % (512 - 16));
+        glTexSubImage2D(GL_TEXTURE_2D, 0, x, y, 16, 16, GL_RGBA, GL_UNSIGNED_BYTE, g_scratch);
+    }
+    for (long i = 0; i < b; ++i) {
+        int slot = (int)(((long)frame * b + i) % MAX_SECTIONS);
+        long bytes = g_quadsPerSection * 4 * VERT_STRIDE;
+        glBindBuffer(GL_ARRAY_BUFFER, g_vbo[slot]);
+        glBufferData(GL_ARRAY_BUFFER, bytes, NULL, GL_STATIC_DRAW);
+        glBufferSubData(GL_ARRAY_BUFFER, 0, bytes, g_scratch);
+    }
+    long entityEvery = 10;
+    for (long i = 0; i < a; ++i) {
+        if (i % entityEvery == entityEvery - 1) {
+            glUseProgram(g_progEntity);
+            glUniformMatrix4fv(g_uMvpEntity, 1, 0, g_mvp);
+            glBindTexture(GL_TEXTURE_2D, g_texEntity);
+            glBindVertexArray(g_vao[i % MAX_SECTIONS]);
+            glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+            glUseProgram(g_progChunk);
+            glBindTexture(GL_TEXTURE_2D, g_texAtlas);
+        } else {
+            glBindVertexArray(g_vao[i % MAX_SECTIONS]);
+            glUniform3f(g_uOffsetChunk, (float)(i & 15), (float)((i >> 4) & 15), 0.0f);
+            glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+        }
+    }
+}
+
+/* ---- Trace-derived cases -------------------------------------------------
+ * Per-frame call mixes measured from the three captured Minecraft traces
+ * (render distance 32, 1280x720, hovering in-world). Each case reproduces one
+ * renderer's dominant per-draw sequence at its measured rate, so the number a
+ * backend posts here is directly comparable to what that game version asks of
+ * the driver every frame.
+ *
+ *   vanilla 1.21.1 : 5495 glDrawElements, 5490 glBindVertexArray,
+ *                    5487 glUniform3fv, 95 glTexSubImage2D (+382 glPixelStorei,
+ *                    247 glTexParameteri), 23 glBufferData per frame
+ *   fabric+sodium  : 132 glMultiDrawElementsBaseVertex, 279 glBindVertexArray,
+ *                    132 glUniform3f, 32 glBufferData per frame
+ *   26.2 snapshot  : 3401 glDrawElementsBaseVertex, each preceded by
+ *                    glBindBufferRange + glBindBuffer (3639/3412 per frame)
+ */
+/* vanilla: bind VAO, push the chunk offset, draw. a = draws per frame. */
+static void case_mc_vanilla_draw(int frame, long a, long b) {
+    (void)frame; (void)b;
+    float offset[3];
+    for (long i = 0; i < a; ++i) {
+        glBindVertexArray(g_vao[i % MAX_SECTIONS]);
+        offset[0] = (float)(i & 15);
+        offset[1] = (float)((i >> 4) & 15);
+        offset[2] = 0.0f;
+        glUniform3fv(g_uOffsetChunk, 1, offset);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+}
+
+/* sodium: one multi-draw covers many chunk sections out of a shared buffer.
+ * a = multi-draws per frame, b = sub-draws inside each. */
+static void case_mc_sodium_multidraw(int frame, long a, long b) {
+    (void)frame;
+    enum { kMaxSub = 64 };
+    if (b <= 0 || b > kMaxSub) b = 32;
+    GLsizei counts[kMaxSub];
+    const void* offsets[kMaxSub];
+    GLint baseVertices[kMaxSub];
+    for (long s = 0; s < b; ++s) {
+        counts[s] = (GLsizei)(g_quadsPerSection * 6 / b);
+        offsets[s] = (const void*)(uintptr_t)(s * (g_quadsPerSection * 6 / b) * 4);
+        baseVertices[s] = 0;
+    }
+    for (long i = 0; i < a; ++i) {
+        glBindVertexArray(g_vao[i % MAX_SECTIONS]);
+        glBindVertexArray(g_vao[i % MAX_SECTIONS]); /* sodium rebinds ~2x per draw */
+        glUniform3f(g_uOffsetChunk, (float)(i & 15), (float)((i >> 4) & 15), 0.0f);
+        // Routed through the includer: GLES has no multi-draw-with-base-vertex, so
+        // a native-driver harness emulates it with the loop the extension folds up.
+        bench_multi_draw_elements_base_vertex(GL_TRIANGLES, counts, GL_UNSIGNED_INT, offsets,
+                                              (GLsizei)b, baseVertices);
+    }
+}
+
+/* 26.2: every draw rebinds a fresh uniform-buffer range out of a ring.
+ * a = draws per frame. */
+static void case_mc_ubo_range(int frame, long a, long b) {
+    (void)b;
+    const size_t slots = (4u * 1024u * 1024u) / g_uboSlot;
+    for (long i = 0; i < a; ++i) {
+        const size_t slot = (size_t)(((long)frame * a + i) % (long)slots);
+        glBindBufferRange(GL_UNIFORM_BUFFER, 0, g_uboRing, (GLintptr)(slot * g_uboSlot),
+                          (GLsizeiptr)g_uboSlot);
+        glBindBuffer(GL_UNIFORM_BUFFER, g_uboRing);
+        glDrawElementsBaseVertex(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0, 0);
+    }
+}
+
+/* vanilla's animated-sprite path: every upload is wrapped in the pixel-store
+ * and filter state Blaze3D re-sets around it. a = uploads per frame. */
+static void case_mc_tex_stream(int frame, long a, long b) {
+    (void)b;
+    glBindVertexArray(g_vao[0]);
+    glBindTexture(GL_TEXTURE_2D, g_texAtlas);
+    for (long i = 0; i < a; ++i) {
+        glPixelStorei(GL_UNPACK_ALIGNMENT, 4);
+        glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
+        glPixelStorei(GL_UNPACK_SKIP_ROWS, 0);
+        glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+        int x = (int)((frame * 13 + i * 17) % (1024 - 16));
+        int y = (int)((frame * 7 + i * 29) % (512 - 16));
+        glTexSubImage2D(GL_TEXTURE_2D, 0, x, y, 16, 16, GL_RGBA, GL_UNSIGNED_BYTE, g_scratch);
+    }
+    glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+}
+
+/* Blaze3D re-resolves uniform locations by name every frame. a = lookups. */
+static void case_mc_uniform_lookup(int frame, long a, long b) {
+    (void)frame; (void)b;
+    static const char* names[4] = {"uMvp", "uOffset", "uAtlas", "uLight"};
+    volatile GLint sink = 0;
+    for (long i = 0; i < a; ++i) sink += glGetUniformLocation(g_progChunk, names[i & 3]);
+    (void)sink;
+    glBindVertexArray(g_vao[0]);
+    glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+}
+
+/* 26.2 rebinds a sampler object per texture unit switch. a = switches. */
+static void case_mc_sampler_churn(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) {
+        glActiveTexture(GL_TEXTURE0 + (GLenum)(i & 3));
+        glBindTexture(GL_TEXTURE_2D, (i & 1) ? g_texEntity : g_texAtlas);
+        glBindSampler((GLuint)(i & 3), g_sampler);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+    glActiveTexture(GL_TEXTURE0);
+}
+
+
+/* 26.2 switches render targets constantly: 132 glBindFramebuffer and 198
+ * glDrawBuffers per frame. Pass switching is where a Vulkan backend pays for
+ * render-pass breaks, so this case is the one to watch on Magma. a = passes. */
+static void case_mc_pass_switch(int frame, long a, long b) {
+    (void)frame; (void)b;
+    static const GLenum kColor0[1] = {GL_COLOR_ATTACHMENT0};
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) {
+        glBindFramebuffer(GL_FRAMEBUFFER, g_passFbo[i & 1]);
+        glDrawBuffers(1, kColor0);
+        glViewport(0, 0, 256, 256);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+    glBindFramebuffer(GL_FRAMEBUFFER, g_mainFbo);
+    glViewport(0, 0, 1280, 720);
+}
+
+/* Blaze3D toggles blend around batches: 46 glEnable/glDisable pairs and 28
+ * glBlendFuncSeparate per vanilla frame. a = toggle pairs. */
+static void case_mc_state_toggle(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) {
+        glEnable(GL_BLEND);
+        glBlendFuncSeparate(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA, GL_ONE, GL_ZERO);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+        glDisable(GL_BLEND);
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+}
+
+/* 26.2 re-sets texture parameters relentlessly - 612 glTexParameteri per frame,
+ * almost always to the value already in place. Measures redundant-param
+ * filtering. a = parameter writes. */
+static void case_mc_tex_param(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    glBindTexture(GL_TEXTURE_2D, g_texAtlas);
+    for (long i = 0; i < a; i += 4) {
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST_MIPMAP_LINEAR);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+    }
+    glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+}
+
+/* Sodium switches programs mid-frame far more than vanilla: 62 glUseProgram and
+ * 60 mat4 uploads per frame. a = program switches. */
+static void case_mc_use_program(int frame, long a, long b) {
+    (void)frame; (void)b;
+    glBindVertexArray(g_vao[0]);
+    for (long i = 0; i < a; ++i) {
+        if (i & 1) {
+            glUseProgram(g_progEntity);
+            glUniformMatrix4fv(g_uMvpEntity, 1, 0, g_mvp);
+        } else {
+            glUseProgram(g_progChunk);
+            glUniformMatrix4fv(g_uMvpChunk, 1, 0, g_mvp);
+        }
+        glDrawElements(GL_TRIANGLES, g_quadsPerSection * 6, GL_UNSIGNED_INT, 0);
+    }
+    glUseProgram(g_progChunk);
+}
+
+/* ---- the case table both harnesses iterate --------------------------------
+ * a/b are the case's own knobs; opsPerFrame is what one bench frame is
+ * normalised by, so ns_per_op compares across renderers. The mc_* rates are
+ * the per-frame call counts measured from the captured traces.
+ */
+typedef void (*bench_case_fn)(int frame, long a, long b);
+
+typedef struct {
+    const char* name;
+    bench_case_fn fn;
+    long a, b, opsPerFrame;
+} BenchCaseDesc;
+
+static const BenchCaseDesc kBenchCases[] = {
+    {"mc_vanilla_draw", case_mc_vanilla_draw, 5495, 0, 5495},
+    {"mc_sodium_multidraw", case_mc_sodium_multidraw, 132, 32, 132},
+    {"mc_ubo_range", case_mc_ubo_range, 3401, 0, 3401},
+    {"mc_tex_stream", case_mc_tex_stream, 95, 0, 95},
+    {"mc_uniform_lookup", case_mc_uniform_lookup, 41, 0, 41},
+    {"mc_sampler_churn", case_mc_sampler_churn, 306, 0, 306},
+    {"mc_pass_switch", case_mc_pass_switch, 132, 0, 132},
+    {"mc_state_toggle", case_mc_state_toggle, 46, 0, 46},
+    {"mc_tex_param", case_mc_tex_param, 612, 0, 612},
+    {"mc_use_program", case_mc_use_program, 62, 0, 62},
+    {"draw_tiny", case_draw_tiny, 2048, 0, 2048},
+    {"draw_uniform", case_draw_uniform, 2048, 0, 2048},
+    {"draw_multi_vao", case_draw_multi_vao, 2048, 0, 2048},
+    {"tex_pingpong", case_tex_pingpong, 1024, 0, 1024},
+    {"program_pingpong", case_program_pingpong, 512, 0, 512},
+    {"chunk_upload", case_chunk_upload, 24, 0, 24},
+    {"atlas_sprite", case_atlas_sprite, 32, 0, 32},
+    {"lightmap", case_lightmap, 4, 0, 4},
+    {"scene_mix", case_scene_mix, 2048, 12, 2048},
+};
+static const int kBenchCaseCount = (int)(sizeof kBenchCases / sizeof kBenchCases[0]);

MobileGL/MG_Benchmark/Driver/run_driver_bench.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+# Run the headless EGL DriverBench on one renderer:
+#   ./run_driver_bench.sh native            [bench args...]
+#   ./run_driver_bench.sh espryt <libMobileGL.so> [bench args...]
+#   ./run_driver_bench.sh magma  <libMobileGL.so> [bench args...]
+# The bench dlopens exactly one EGL provider (DRIVERBENCH_EGL_LIB): the system
+# libEGL.so.1 for native, or the given libMobileGL.so for a MobileGL backend -
+# no LD_LIBRARY_PATH shadowing, so MobileGL's own loader still finds the real
+# driver underneath.
+#
+# Pin the vendor libraries explicitly. A bare libEGL.so.1 on a glvnd system
+# picks whatever vendor eglGetDisplay(EGL_DEFAULT_DISPLAY) resolves first,
+# which is Mesa/llvmpipe here - a software rasteriser silently replacing the
+# GPU under a benchmark. Override MGL_EGL_VENDOR / MGL_VK_ICD to test another
+# driver.
+set -eu
+HERE=$(cd "$(dirname "$0")" && pwd)
+BENCH=${DRIVERBENCH_BIN:-$HERE/DriverBench}
+EGL_VENDOR=${MGL_EGL_VENDOR:-/usr/share/glvnd/egl_vendor.d/10_nvidia.json}
+VK_ICD=${MGL_VK_ICD:-/usr/share/vulkan/icd.d/nvidia_icd.x86_64.json}
+MODE=$1; shift
+
+export __EGL_VENDOR_LIBRARY_FILENAMES=$EGL_VENDOR
+export EGL_PLATFORM=${EGL_PLATFORM:-x11}
+
+case "$MODE" in
+  native)
+    export DRIVERBENCH_EGL_LIB=${DRIVERBENCH_EGL_LIB:-libEGL.so.1}
+    ;;
+  espryt)
+    export DRIVERBENCH_EGL_LIB=$(readlink -f "$1"); shift
+    export MOBILEGL_BACKEND_TYPE=DirectGLES
+    ;;
+  magma)
+    export DRIVERBENCH_EGL_LIB=$(readlink -f "$1"); shift
+    export MOBILEGL_BACKEND_TYPE=DirectVulkan
+    export VK_ICD_FILENAMES=$VK_ICD
+    ;;
+  *) echo "unknown mode: $MODE (native|espryt|magma)"; exit 1 ;;
+esac
+exec "$BENCH" "$@"

MobileGL/MG_Impl/EGLImpl/EGLImpl.cpp

@@ -21,7 +21,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
 
         EGLStateContext* GetState() {
             if (!MG_State::pEGLContext) {
-                MGLOG_E("pEGLContext is null. MG_State may not be initialized.");
+                MGLOG_E_ONCE("pEGLContext is null. MG_State may not be initialized.");
             }
             return MG_State::pEGLContext.get();
         }
@@ -146,7 +146,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
 
         auto* backendObject = GetBackendObject(state);
         if (!backendObject) {
-            MGLOG_E("activeBackendObject not initialized!");
+            MGLOG_E_ONCE("activeBackendObject not initialized!");
             state->DestroySurface(dpy, surface);
             return EGL_NO_SURFACE;
         }
@@ -172,11 +172,11 @@ namespace MobileGL::MG_Impl::EGLImpl {
 
         auto* backendObject = GetBackendObject(state);
         if (!backendObject) {
-            MGLOG_E("activeBackendObject not initialized!");
+            MGLOG_E_ONCE("activeBackendObject not initialized!");
             return EGL_FALSE;
         }
         if (!backendObject->SwapEGLBuffers(dpy, draw)) {
-            MGLOG_E("eglSwapBuffers failed on thread=%s dpy=%p draw=%p", CurrentThreadIdString().c_str(), dpy, draw);
+            MGLOG_E_ONCE("eglSwapBuffers failed on thread=%s dpy=%p draw=%p", CurrentThreadIdString().c_str(), dpy, draw);
             state->SetError(EGL_BAD_SURFACE);
             return EGL_FALSE;
         }
@@ -211,7 +211,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
 
         auto* backendObject = GetBackendObject(state);
         if (!backendObject) {
-            MGLOG_E("activeBackendObject not initialized!");
+            MGLOG_E_ONCE("activeBackendObject not initialized!");
             return EGL_FALSE;
         }
         if (!backendObject->InitializeEGLDisplay(dpy, major, minor)) {
@@ -265,7 +265,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
         if (releaseCurrentRequest) {
             if (auto* backendObject = MG_Backend::pActiveBackendObject.get()) {
                 if (!backendObject->MakeEGLCurrent(dpy, draw, read, ctx)) {
-                    MGLOG_E("eglMakeCurrent release failed in backend thread=%s", threadId.c_str());
+                    MGLOG_E_ONCE("eglMakeCurrent release failed in backend thread=%s", threadId.c_str());
                     state->MakeCurrent(oldDisplay, oldDraw, oldRead, oldContext);
                     state->SetError(EGL_BAD_ACCESS);
                     return EGL_FALSE;
@@ -277,12 +277,12 @@ namespace MobileGL::MG_Impl::EGLImpl {
 
         auto* backendObject = GetBackendObject(state);
         if (!backendObject) {
-            MGLOG_E("activeBackendObject not initialized!");
+            MGLOG_E_ONCE("activeBackendObject not initialized!");
             state->MakeCurrent(oldDisplay, oldDraw, oldRead, oldContext);
             return EGL_FALSE;
         }
         if (!backendObject->MakeEGLCurrent(dpy, draw, read, ctx)) {
-            MGLOG_E("eglMakeCurrent backend attach failed thread=%s dpy=%p draw=%p read=%p ctx=%p", threadId.c_str(),
+            MGLOG_E_ONCE("eglMakeCurrent backend attach failed thread=%s dpy=%p draw=%p read=%p ctx=%p", threadId.c_str(),
                     dpy, draw, read, ctx);
             state->SetError(EGL_BAD_ACCESS);
             state->MakeCurrent(oldDisplay, oldDraw, oldRead, oldContext);
@@ -703,7 +703,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
 
         auto* backendObject = GetBackendObject(state);
         if (!backendObject) {
-            MGLOG_E("activeBackendObject not initialized!");
+            MGLOG_E_ONCE("activeBackendObject not initialized!");
             state->DestroySurface(dpy, surface);
             return EGL_NO_SURFACE;
         }
@@ -726,7 +726,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
         }
         auto* backendObject = GetBackendObject(state);
         if (!backendObject) {
-            MGLOG_E("activeBackendObject not initialized!");
+            MGLOG_E_ONCE("activeBackendObject not initialized!");
             return EGL_FALSE;
         }
         width = std::max<EGLint>(width, 1);
@@ -764,7 +764,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
         MGLOG_D("eglGetProcAddress(%s)", name);
         void* proc = MG_Impl::GetProcAddress(name);
         if (!proc) {
-            MGLOG_W("Failed to get function: %s", name);
+            MGLOG_D("Failed to get function: %s", name);
             return nullptr;
         }
         return (__eglMustCastToProperFunctionPointerType)proc;

MobileGL/MG_Impl/GLImpl/Buffer/GL_Buffer.cpp

@@ -9,6 +9,7 @@
 #include "GL_Buffer.h"
 #include "Validators.h"
 #include "../Texture/GL_Texture.h"
+#include "../Getter/GL_Getter.h"
 #include <MG_Util/Converters/GLToMG/TextureEnumConverter.h>
 #include <MG_Util/Metrics/TextureMetrics.h>
 #include <Config.h>
@@ -17,6 +18,7 @@
 #include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
 #include <MG_Util/Converters/GLToMG/BufferEnumConverter.h>
 #include <MG_Util/Converters/MGToGL/BufferEnumConverter.h>
+#include <MG_Util/Texture/PixelStoreProcessor.h>
 
 namespace MobileGL::MG_Impl::GLImpl {
     namespace {
@@ -30,6 +32,8 @@ namespace MobileGL::MG_Impl::GLImpl {
             NamedBufferData,
             NamedBufferSubData,
             CopyNamedBufferSubData,
+            ClearBufferData,
+            ClearBufferSubData,
             ClearNamedBufferData,
             ClearNamedBufferSubData,
             MapBufferRange,
@@ -64,6 +68,10 @@ namespace MobileGL::MG_Impl::GLImpl {
                 return "NamedBufferSubData";
             case BufferOp::CopyNamedBufferSubData:
                 return "CopyNamedBufferSubData";
+            case BufferOp::ClearBufferData:
+                return "ClearBufferData";
+            case BufferOp::ClearBufferSubData:
+                return "ClearBufferSubData";
             case BufferOp::ClearNamedBufferData:
                 return "ClearNamedBufferData";
             case BufferOp::ClearNamedBufferSubData:
@@ -142,16 +150,6 @@ namespace MobileGL::MG_Impl::GLImpl {
                 return 0;
             }
 
-            // The pattern is replicated verbatim, which is only the whole story while the client
-            // layout already matches the internal format - the case every entry point in practice
-            // uses, and the only one the conversion machinery here can express. Say so rather than
-            // quietly writing a differently-sized pattern.
-            const SizeT sourceSize = MG_Util::GetInputBytesPerPixel(inputFormat, pixelType);
-            if (sourceSize != elementSize) {
-                MGLOG_W("%s: clear pattern is %zu bytes but internalformat 0x%X stores %zu; "
-                        "converting between them is not implemented",
-                        GetBufferOpName(op), sourceSize, internalformat, elementSize);
-            }
             return elementSize;
         }
 
@@ -193,27 +191,59 @@ namespace MobileGL::MG_Impl::GLImpl {
             return true;
         }
 
-        void ClearNamedBufferRange_State(GLuint buffer, GLenum internalformat, GLintptr offset, GLsizeiptr size,
+        Bool BuildClearPattern(GLenum internalformat, GLenum format, GLenum type, const void* data,
+                               SizeT patternSize, BufferOp op, Vector<Uint8>& pattern) {
+            const TextureInternalFormat internal = MG_Util::ConvertGLEnumToTextureInternalFormat(internalformat);
+            const TextureInputFormat inputFormat = MG_Util::ConvertGLEnumToTextureInputFormat(format);
+            const TexturePixelDataType inputType = MG_Util::ConvertGLEnumToTexturePixelDataType(type);
+
+            Vector<Uint8> zeroInput;
+            const void* inputPixel = data;
+            if (inputPixel == nullptr) {
+                const SizeT inputSize = MG_Util::GetInputBytesPerPixel(inputFormat, inputType);
+                if (inputSize == 0) {
+                    MG_State::pGLContext->RecordError(
+                        ErrorCode::InvalidValue,
+                        MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", GetBufferOpName(op),
+                                                     "format and type do not describe a source pixel."));
+                    return false;
+                }
+                zeroInput.resize(inputSize);
+                inputPixel = zeroInput.data();
+            }
+
+            if (!MG_Util::PixelStoreProcessor::ConvertOnePixelToInternal(
+                    internal, inputFormat, inputType, inputPixel, pattern)) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>(
+                        "MG_Impl/GLImpl", GetBufferOpName(op),
+                        std::format("Cannot convert one ({}, {}) pixel into internalformat 0x{:X}.",
+                                    MG_Util::ConvertGLEnumToString(format), MG_Util::ConvertGLEnumToString(type),
+                                    internalformat)));
+                return false;
+            }
+
+            if (data == nullptr) {
+                // GL defines a null clear value as all zero bits in the destination store, while
+                // retaining the format/type validation above.
+                pattern.assign(patternSize, 0);
+            }
+            return true;
+        }
+
+        void ClearBufferRange_State(const SharedPtr<MG_State::GLState::BufferObject>& bufferObject,
+                                    GLenum internalformat, GLintptr offset, GLsizeiptr size,
                                     GLenum format, GLenum type, const void* data, BufferOp op) {
             const SizeT patternSize = GetClearPatternSize(internalformat, format, type, op);
             if (patternSize == 0) return;
-
-            auto bufferObject = GetNamedBufferObject(buffer, op);
-            if (!bufferObject) return;
             if (!ValidateBufferClearRange(bufferObject, offset, size, patternSize, op)) return;
             if (size == 0) return;
 
-            Vector<Uint8> clearData(static_cast<SizeT>(size));
+            Vector<Uint8> pattern;
-            if (data) {
+            if (!BuildClearPattern(internalformat, format, type, data, patternSize, op, pattern)) return;
-                const auto* pattern = static_cast<const Uint8*>(data);
+            bufferObject->FillSubData({pattern.data(), pattern.size()}, static_cast<SizeT>(offset),
-                for (SizeT at = 0; at < clearData.size(); at += patternSize) {
+                                      static_cast<SizeT>(size));
-                    Memcpy(clearData.data() + at, pattern, patternSize);
-                }
-            } else {
-                Memset(clearData.data(), 0, clearData.size());
-            }
-
-            bufferObject->UploadSubData({clearData.data(), clearData.size()}, static_cast<SizeT>(offset));
         }
 
         auto& GetBufferBindingSlot(BufferTarget target) {
@@ -861,6 +891,10 @@ namespace MobileGL::MG_Impl::GLImpl {
 
         Range1D mappedRange = bufferObject->GetMappedRange();
         auto mappingAccess = bufferObject->GetMappingAccess();
+        // GL 4.6 6.5: the error is on OVERLAP with the mapped range, i.e. a half-open
+        // intersection test. There used to be a second test below this one asking only
+        // `offset + size >= mappedRange.start`, which rejects every write that starts
+        // before a mapped tail as well - it made a legal disjoint glBufferSubData fail.
         if (bufferObject->IsMapped() && !(mappingAccess & BufferMappingAccessBit::Persistent) &&
             (offset < mappedRange.end) && (offset + size > mappedRange.start)) {
             MG_State::pGLContext->RecordError(
@@ -871,18 +905,6 @@ namespace MobileGL::MG_Impl::GLImpl {
             return;
         }
 
-        if (bufferObject->IsMapped() && !(mappingAccess & BufferMappingAccessBit::Persistent)) {
-            Range1D mappedRange = bufferObject->GetMappedRange();
-            if (offset + size >= mappedRange.start) {
-                MG_State::pGLContext->RecordError(
-                    ErrorCode::InvalidOperation,
-                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "BufferSubData_State",
-                                                 "Cannot modify a mapped buffer object unless it was "
-                                                 "mapped with GL_MAP_PERSISTENT_BIT."));
-                return;
-            }
-        }
-
         bufferObject->UploadSubData({(void*)data, (SizeT)size}, offset);
     }
 
@@ -1013,6 +1035,11 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void BufferStorage_State(GLenum target, GLsizeiptr size, const void* data, GLbitfield flags) {
+        // Error precedence: "no buffer is bound to target" outranks a bad size or bad
+        // flags, so the binding has to be resolved before either is validated.
+        auto bufferObject = GetBoundBufferObject(target, BufferOp::BufferStorage);
+        if (!bufferObject) return;
+
         if (size <= 0) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue,
@@ -1021,8 +1048,6 @@ namespace MobileGL::MG_Impl::GLImpl {
         }
         if (!ValidateStorageFlags(flags, BufferOp::BufferStorage)) return;
 
-        auto bufferObject = GetBoundBufferObject(target, BufferOp::BufferStorage);
-        if (!bufferObject) return;
         if (bufferObject->IsImmutableStorage()) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
@@ -1057,6 +1082,10 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void NamedBufferStorage_State(GLuint buffer, GLsizeiptr size, const void* data, GLbitfield flags) {
+        // Same precedence as BufferStorage_State: the buffer-name error comes first.
+        auto bufferObject = GetNamedBufferObject(buffer, BufferOp::NamedBufferStorage);
+        if (!bufferObject) return;
+
         if (size <= 0) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue,
@@ -1065,8 +1094,6 @@ namespace MobileGL::MG_Impl::GLImpl {
         }
         if (!ValidateStorageFlags(flags, BufferOp::NamedBufferStorage)) return;
 
-        auto bufferObject = GetNamedBufferObject(buffer, BufferOp::NamedBufferStorage);
-        if (!bufferObject) return;
         if (bufferObject->IsImmutableStorage()) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
@@ -1199,16 +1226,33 @@ namespace MobileGL::MG_Impl::GLImpl {
                                         static_cast<SizeT>(writeOffset), static_cast<SizeT>(size));
     }
 
+    void ClearBufferData_State(GLenum target, GLenum internalformat, GLenum format, GLenum type, const void* data) {
+        auto bufferObject = GetBoundBufferObject(target, BufferOp::ClearBufferData);
+        if (!bufferObject) return;
+        ClearBufferRange_State(bufferObject, internalformat, 0, static_cast<GLsizeiptr>(bufferObject->GetSize()), format,
+                               type, data, BufferOp::ClearBufferData);
+    }
+
+    void ClearBufferSubData_State(GLenum target, GLenum internalformat, GLintptr offset, GLsizeiptr size,
+                                  GLenum format, GLenum type, const void* data) {
+        auto bufferObject = GetBoundBufferObject(target, BufferOp::ClearBufferSubData);
+        if (!bufferObject) return;
+        ClearBufferRange_State(bufferObject, internalformat, offset, size, format, type, data,
+                               BufferOp::ClearBufferSubData);
+    }
+
     void ClearNamedBufferData_State(GLuint buffer, GLenum internalformat, GLenum format, GLenum type, const void* data) {
         auto bufferObject = GetNamedBufferObject(buffer, BufferOp::ClearNamedBufferData);
         if (!bufferObject) return;
-        ClearNamedBufferRange_State(buffer, internalformat, 0, static_cast<GLsizeiptr>(bufferObject->GetSize()), format,
+        ClearBufferRange_State(bufferObject, internalformat, 0, static_cast<GLsizeiptr>(bufferObject->GetSize()), format,
                                type, data, BufferOp::ClearNamedBufferData);
     }
 
     void ClearNamedBufferSubData_State(GLuint buffer, GLenum internalformat, GLintptr offset, GLsizeiptr size,
                                        GLenum format, GLenum type, const void* data) {
-        ClearNamedBufferRange_State(buffer, internalformat, offset, size, format, type, data,
+        auto bufferObject = GetNamedBufferObject(buffer, BufferOp::ClearNamedBufferSubData);
+        if (!bufferObject) return;
+        ClearBufferRange_State(bufferObject, internalformat, offset, size, format, type, data,
                                BufferOp::ClearNamedBufferSubData);
     }
 
@@ -1486,12 +1530,87 @@ namespace MobileGL::MG_Impl::GLImpl {
         GetBufferBindingSlot(bufferTarget).Bind(bufferObject);
     }
 
+    // GL 4.6 core 6.1.1: the constraints glBindBufferRange puts on the (offset, size) pair.
+    // Every one of them is INVALID_VALUE, and all of them are checked before a single piece
+    // of state is written - a rejected bind must leave the binding point exactly as it was.
+    // They apply only to a non-zero buffer: buffer 0 detaches the binding point and ignores
+    // offset and size, which is also how glBindBuffersRange spells "reset this element"
+    // (a NULL buffers array, or a zero entry inside one).
+    static Bool ValidateBufferRangeOffsetAndSize(GLenum target, GLintptr offset, GLsizeiptr size,
+                                                 const char* funcName, Bool hasBuffer = true) {
+        if (hasBuffer && size <= 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", funcName,
+                                             std::format("size ({}) must be greater than zero.", size)));
+            return false;
+        }
+        if (offset < 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", funcName,
+                                             std::format("offset ({}) must not be negative.", offset)));
+            return false;
+        }
+        // GL_UNIFORM_BUFFER and GL_SHADER_STORAGE_BUFFER each constrain the offset to their own
+        // implementation-defined alignment, which glGetIntegerv already answers.
+        GLenum alignmentQuery = GL_NONE;
+        if (target == GL_SHADER_STORAGE_BUFFER) {
+            alignmentQuery = GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT;
+        } else if (target == GL_UNIFORM_BUFFER) {
+            alignmentQuery = GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT;
+        }
+        if (alignmentQuery != GL_NONE) {
+            GLint alignment = 0;
+            GetIntegerv(alignmentQuery, &alignment);
+            if (alignment > 0 && (offset % static_cast<GLintptr>(alignment)) != 0) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>(
+                        "MG_Impl/GLImpl", funcName,
+                        std::format("offset ({}) must be a multiple of {} ({}).", offset,
+                                    MG_Util::ConvertGLEnumToString(alignmentQuery), alignment)));
+                return false;
+            }
+        }
+        // GL 4.6 core 6.1.1 constrains the OFFSET to a multiple of four for both
+        // TRANSFORM_FEEDBACK_BUFFER and ATOMIC_COUNTER_BUFFER (the atomic-counter one has no
+        // queryable alignment pname, which is why it was missing here), and the SIZE only for
+        // transform feedback, whose capture is written in whole 32-bit components. Extending the
+        // size rule to atomic counters as well breaks a legal bind: the conformance suite splits
+        // MAX_ATOMIC_COUNTER_BUFFER_SIZE evenly across the binding points and that quotient is
+        // not required to land on four.
+        if ((target == GL_TRANSFORM_FEEDBACK_BUFFER || target == GL_ATOMIC_COUNTER_BUFFER) && (offset % 4) != 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", funcName,
+                                             std::format("offset ({}) must be a multiple of 4 for {}.", offset,
+                                                         MG_Util::ConvertGLEnumToString(target))));
+            return false;
+        }
+        if (target == GL_TRANSFORM_FEEDBACK_BUFFER && hasBuffer && (size % 4) != 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl", funcName,
+                    std::format("size ({}) must be a multiple of 4 for GL_TRANSFORM_FEEDBACK_BUFFER.", size)));
+            return false;
+        }
+        return true;
+    }
+
     void BindBufferRange_State(GLenum target, GLuint index, GLuint buffer, GLintptr offset, GLsizeiptr size) {
         MGLOG_D("%s: target = %s, index = %u, buffer = %u, offset = %d, size = %d", __func__,
                 MG_Util::ConvertGLEnumToString(target).c_str(), index, buffer, offset, size);
         BufferTarget bufferTarget = MG_Util::ConvertGLEnumToBufferTarget(target);
         if (!BufferImpl::ValidateBufferBindingPointTarget(bufferTarget)) return;
         if (!BufferImpl::ValidateBufferBindingPointIndex(bufferTarget, index)) return;
+        // The target's alignment rules are a property of the BINDING POINT, not of the buffer,
+        // so they apply even when buffer is zero - which is exactly how
+        // KHR-GL43.shader_storage_buffer_object.negative-api-bind probes the SSBO alignment
+        // (glBindBufferRange(SHADER_STORAGE_BUFFER, 0, 0, alignment - 1, 0)). Only the size
+        // rules need a buffer, since buffer 0 detaches the binding point and ignores size.
+        if (!ValidateBufferRangeOffsetAndSize(target, offset, size, __func__, /*hasBuffer: */ buffer != 0)) return;
         if (bufferTarget == BufferTarget::TransformFeedback && MG_State::pGLContext->IsTransformFeedbackActive()) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
@@ -1589,6 +1708,15 @@ namespace MobileGL::MG_Impl::GLImpl {
         CopyNamedBufferSubData_State(readBuffer, writeBuffer, readOffset, writeOffset, size);
     }
 
+    void ClearBufferData(GLenum target, GLenum internalformat, GLenum format, GLenum type, const void* data) {
+        ClearBufferData_State(target, internalformat, format, type, data);
+    }
+
+    void ClearBufferSubData(GLenum target, GLenum internalformat, GLintptr offset, GLsizeiptr size, GLenum format,
+                            GLenum type, const void* data) {
+        ClearBufferSubData_State(target, internalformat, offset, size, format, type, data);
+    }
+
     void ClearNamedBufferData(GLuint buffer, GLenum internalformat, GLenum format, GLenum type, const void* data) {
         ClearNamedBufferData_State(buffer, internalformat, format, type, data);
     }
@@ -1665,16 +1793,54 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     // ARB_multi_bind: defined by the spec as equivalent to a loop over the single-bind entry
-    // points (with buffer 0 resetting the binding point).
+    // points (with buffer 0 resetting the binding point) - but only AFTER an up-front check
+    // of the whole [first, first + count) range. Looping straight into the single-bind entry
+    // points reports the single-bind INVALID_VALUE for an out-of-range index instead of the
+    // multi-bind INVALID_OPERATION, and binds the in-range prefix before failing.
+    static Bool ValidateMultiBindBufferRange(GLenum target, GLuint first, GLsizei count, const char* funcName) {
+        BufferTarget bufferTarget = MG_Util::ConvertGLEnumToBufferTarget(target);
+        if (!BufferImpl::ValidateBufferBindingPointTarget(bufferTarget)) return false;
+        return BufferImpl::ValidateBufferBindingPointRange(bufferTarget, first, count, funcName);
+    }
+
+    // ARB_multi_bind states the equivalence to a loop of single binds "except that ... buffers
+    // will not be created if they do not exist": glBindBuffer instantiates a name glGenBuffers
+    // merely reserved, glBindBuffers* must refuse it and raise INVALID_OPERATION instead
+    // (KHR-GL44.multi_bind.errors_bind_buffers).
+    //
+    // Deliberately PER ELEMENT, not all-or-nothing: the equivalence the extension defines is a
+    // loop, so a bad entry costs its own binding point and nothing else. Rejecting the whole
+    // call instead cost multi_bind.functional_bind_buffers_base its bindings.
+    static Bool IsExistingBufferForMultiBind(GLuint buffer, GLsizei index, const char* funcName) {
+        if (buffer == 0 || MG_State::pGLContext->ValidateBufferObject(buffer)) return true;
+        MG_State::pGLContext->RecordError(
+            ErrorCode::InvalidOperation,
+            MakeUnique<GenericErrorInfo>(
+                "MG_Impl/GLImpl", funcName,
+                std::format("buffers[{}] ({}) is not the name of an existing buffer object.", index, buffer)));
+        return false;
+    }
+
     void BindBuffersBase(GLenum target, GLuint first, GLsizei count, const GLuint* buffers) {
+        if (!ValidateMultiBindBufferRange(target, first, count, __func__)) return;
         for (GLsizei i = 0; i < count; ++i) {
-            BindBufferBase_State(target, first + i, buffers ? buffers[i] : 0);
+            const GLuint buffer = buffers ? buffers[i] : 0;
+            if (!IsExistingBufferForMultiBind(buffer, i, __func__)) continue;
+            BindBufferBase_State(target, first + i, buffer);
         }
     }
 
+    // The (offset, size) constraints are the one part of glBindBuffersRange that stays
+    // per-element: ARB_multi_bind checks them separately for each binding point, leaves that
+    // point unchanged on failure, and still applies the remaining elements - which is exactly
+    // what looping into BindBufferRange_State does. Only the [first, first + count) range is
+    // an up-front, all-or-nothing check. Elements that name buffer 0 (or a NULL buffers array)
+    // reset the binding point through BindBufferBase_State and carry no offset/size to check.
     void BindBuffersRange(GLenum target, GLuint first, GLsizei count, const GLuint* buffers, const GLintptr* offsets,
                           const GLsizeiptr* sizes) {
+        if (!ValidateMultiBindBufferRange(target, first, count, __func__)) return;
         for (GLsizei i = 0; i < count; ++i) {
+            if (buffers && !IsExistingBufferForMultiBind(buffers[i], i, __func__)) continue;
             if (!buffers || buffers[i] == 0) {
                 BindBufferBase_State(target, first + i, 0);
             } else {

MobileGL/MG_Impl/GLImpl/Buffer/GL_Buffer.h

@@ -27,6 +27,9 @@ namespace MobileGL::MG_Impl::GLImpl {
     void NamedBufferSubData(GLuint buffer, GLintptr offset, GLsizeiptr size, const void* data);
     void CopyNamedBufferSubData(GLuint readBuffer, GLuint writeBuffer, GLintptr readOffset, GLintptr writeOffset,
                                 GLsizeiptr size);
+    void ClearBufferData(GLenum target, GLenum internalformat, GLenum format, GLenum type, const void* data);
+    void ClearBufferSubData(GLenum target, GLenum internalformat, GLintptr offset, GLsizeiptr size, GLenum format,
+                            GLenum type, const void* data);
     void ClearNamedBufferData(GLuint buffer, GLenum internalformat, GLenum format, GLenum type, const void* data);
     void ClearNamedBufferSubData(GLuint buffer, GLenum internalformat, GLintptr offset, GLsizeiptr size, GLenum format,
                                  GLenum type, const void* data);

MobileGL/MG_Impl/GLImpl/Buffer/Validators.cpp

@@ -53,7 +53,9 @@ namespace MobileGL::MG_Impl::GLImpl::BufferImpl {
         return true;
     }
 
-    Bool ValidateBufferBindingPointIndex(BufferTarget target, Uint index) {
+    namespace {
+        // The GL-visible number of indexed binding points for `target`.
+        SizeT GetBufferBindingPointLimit(BufferTarget target) {
             SizeT pointCount = MG_State::pGLContext->GetBufferBindingPointCount(target);
             if (target == BufferTarget::ShaderStorage && MG_Backend::pActiveBackendObject) {
                 const Int backendCount =
@@ -65,6 +67,32 @@ namespace MobileGL::MG_Impl::GLImpl::BufferImpl {
                 // binding points in GL 3.3 (no ARB_transform_feedback3).
                 pointCount = std::min<SizeT>(pointCount, 4);
             }
+            return pointCount;
+        }
+    } // namespace
+
+    Bool ValidateBufferBindingPointRange(BufferTarget target, Uint first, GLsizei count, const char* funcName) {
+        if (count < 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue, MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl/BufferImpl", funcName,
+                                                                      "count must be non-negative."));
+            return false;
+        }
+        const SizeT pointCount = GetBufferBindingPointLimit(target);
+        if (static_cast<Uint64>(first) + static_cast<Uint64>(count) > static_cast<Uint64>(pointCount)) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl/BufferImpl", funcName,
+                    std::format("first + count ({} + {}) exceeds the {} indexed binding points of target {}.", first,
+                                count, pointCount, MG_Util::ConvertBufferTargetToString(target))));
+            return false;
+        }
+        return true;
+    }
+
+    Bool ValidateBufferBindingPointIndex(BufferTarget target, Uint index) {
+        const SizeT pointCount = GetBufferBindingPointLimit(target);
 
         if (index < pointCount) {
             return true;

MobileGL/MG_Impl/GLImpl/Buffer/Validators.h

@@ -17,4 +17,8 @@ namespace MobileGL::MG_Impl::GLImpl::BufferImpl {
     Bool ValidateBufferMappingAccess(Flags<BufferMappingAccessBit> accessBits);
     Bool ValidateBufferBindingPointTarget(BufferTarget target);
     Bool ValidateBufferBindingPointIndex(BufferTarget target, Uint index);
+    // ARB_multi_bind: glBindBuffersBase/Range validate the whole [first, first + count) range
+    // up front and report INVALID_OPERATION, where a single out-of-range index would be
+    // INVALID_VALUE. Naively looping the single-bind entry points reports the wrong class.
+    Bool ValidateBufferBindingPointRange(BufferTarget target, Uint first, GLsizei count, const char* funcName);
 } // namespace MobileGL::MG_Impl::GLImpl::BufferImpl

MobileGL/MG_Impl/GLImpl/Drawing/GL_Drawing.cpp

@@ -14,8 +14,8 @@
 #include "../Getter/GL_Getter.h"
 
 namespace MobileGL::MG_Impl::GLImpl {
-    static Bool ValidateCurrentProgramForExecution(const char* functionName) {
+    static Bool ValidateProgramForExecution(const SharedPtr<MG_State::GLState::ProgramObject>& currentProgram,
-        const auto& currentProgram = MG_State::pGLContext->GetCurrentProgram();
+                                            const char* functionName) {
         if (!currentProgram) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
@@ -34,10 +34,17 @@ namespace MobileGL::MG_Impl::GLImpl {
         return true;
     }
 
+    static Bool ValidateCurrentProgramForExecution(const char* functionName) {
+        return ValidateProgramForExecution(MG_State::pGLContext->GetProgramForDraw(), functionName);
+    }
+
+    // A dispatch resolves its program through the DISPATCH accessor: with a pipeline bound
+    // that is the pipeline's compute stage program, not the graphics composite a draw would
+    // build - which no longer contains a compute stage to find at all.
     static Bool ValidateCurrentProgramForCompute(const char* functionName) {
-        if (!ValidateCurrentProgramForExecution(functionName)) return false;
+        const auto& currentProgram = MG_State::pGLContext->GetProgramForDispatch();
+        if (!ValidateProgramForExecution(currentProgram, functionName)) return false;
 
-        const auto& currentProgram = MG_State::pGLContext->GetCurrentProgram();
         if (currentProgram->GetShaderIndexByStage(ShaderStage::Compute) < 0) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
@@ -173,7 +180,7 @@ namespace MobileGL::MG_Impl::GLImpl {
         // input primitive (GL 4.6 core 11.3.1); anything else is INVALID_OPERATION. GL_PATCHES
         // is the tessellation pipeline's input and reaches the geometry stage already
         // converted, so it is not constrained here.
-        const auto& currentProgram = MG_State::pGLContext->GetCurrentProgram();
+        const auto& currentProgram = MG_State::pGLContext->GetProgramForDraw();
         const GLenum gsInput = currentProgram ? currentProgram->GetGeometryInputType() : GL_NONE;
         if (gsInput != GL_NONE && mode != GL_PATCHES) {
             Bool compatible = false;
@@ -474,10 +481,63 @@ namespace MobileGL::MG_Impl::GLImpl {
             return;
         }
         if (!ValidateCurrentProgramForCompute(__func__)) return;
+        // GL 4.6 core 19: each num_groups_* must be within GL_MAX_COMPUTE_WORK_GROUP_COUNT
+        // for its dimension. GetIntegeri_v already floors that at the spec minimum.
+        const GLuint numGroups[3] = {numGroupsX, numGroupsY, numGroupsZ};
+        for (GLuint dimension = 0; dimension < 3; ++dimension) {
+            GLint maxGroups = 0;
+            GetIntegeri_v(GL_MAX_COMPUTE_WORK_GROUP_COUNT, dimension, &maxGroups);
+            if (numGroups[dimension] > static_cast<GLuint>(std::max(maxGroups, 0))) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                                 "num_groups exceeds GL_MAX_COMPUTE_WORK_GROUP_COUNT for dimension " +
+                                                     std::to_string(dimension) + "."));
+                return;
+            }
+        }
         dispatchCompute(numGroupsX, numGroupsY, numGroupsZ);
     }
 
     void DispatchComputeIndirect(GLintptr indirect) {
+        // Argument and binding validation runs FIRST. Both are properties of the call and of GL
+        // state, so a context whose backend cannot dispatch at all must still report the
+        // argument error the spec names rather than masking every one of them with
+        // "unsupported" - which is what put GL_INVALID_OPERATION where
+        // KHR-GL43.compute_shader.api-indirect expects GL_INVALID_VALUE.
+        //
+        // GL 4.6 core 19: `indirect` is a byte offset into GL_DISPATCH_INDIRECT_BUFFER -
+        // negative or misaligned is INVALID_VALUE, nothing bound is INVALID_OPERATION.
+        if (indirect < 0 || (indirect % 4) != 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "indirect must be non-negative and a multiple of 4."));
+            return;
+        }
+        const auto& indirectBuffer =
+            MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DispatchIndirect).GetBoundObject();
+        if (!indirectBuffer) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "No buffer is bound to GL_DISPATCH_INDIRECT_BUFFER."));
+            return;
+        }
+        // ...and the same INVALID_OPERATION covers "the command would source data beyond the end
+        // of the bound buffer object" (GL 4.6 core 19): the dispatch reads three uints starting
+        // at `indirect`.
+        constexpr SizeT kDispatchIndirectCommandSize = 3 * sizeof(Uint32);
+        if (static_cast<SizeT>(indirect) + kDispatchIndirectCommandSize > indirectBuffer->GetSize()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl", __func__,
+                    std::format("indirect ({}) + 12 bytes runs past the end of the {}-byte buffer bound to "
+                                "GL_DISPATCH_INDIRECT_BUFFER.",
+                                indirect, indirectBuffer->GetSize())));
+            return;
+        }
         auto dispatchComputeIndirect = MG_Backend::gBackendFunctionsTable.GL.DispatchComputeIndirect;
         if (!dispatchComputeIndirect) {
             MG_State::pGLContext->RecordError(
@@ -547,8 +607,80 @@ namespace MobileGL::MG_Impl::GLImpl {
         MultiDrawArraysIndirect_Backend(mode, indirect, drawcount, stride);
     }
 
+    // ARB_indirect_parameters / GL 4.6 core 10.4: `drawcount` is a byte offset into the buffer
+    // bound to PARAMETER_BUFFER and holds one uint draw count. Three errors have to be raised
+    // before the call reaches a backend, and none of them was
+    // (KHR-GL46.indirect_parameters_tests.MultiDraw{Arrays,Elements}IndirectCount):
+    //   * drawcount not a multiple of four                                  INVALID_VALUE
+    //   * nothing bound to PARAMETER_BUFFER, or the uint at `drawcount`
+    //     lies past its end                                                 INVALID_OPERATION
+    //   * maxdrawcount commands from `indirect` run past the end of the
+    //     buffer bound to DRAW_INDIRECT_BUFFER                              INVALID_OPERATION
+    static Bool ValidateIndirectCountDraw(GLintptr indirect, GLintptr drawcount, GLsizei maxdrawcount,
+                                          GLsizei stride, SizeT commandSize, const char* funcName) {
+        if (drawcount < 0 || (drawcount % 4) != 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", funcName,
+                                             "drawcount must be non-negative and a multiple of four."));
+            return false;
+        }
+        const auto& parameterBuffer =
+            MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::Parameter).GetBoundObject();
+        if (!parameterBuffer ||
+            static_cast<SizeT>(drawcount) + sizeof(Uint32) > parameterBuffer->GetSize()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", funcName,
+                                             "No buffer is bound to GL_PARAMETER_BUFFER, or drawcount runs past "
+                                             "the end of the one that is."));
+            return false;
+        }
+        if (maxdrawcount < 0 || stride < 0 || indirect < 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", funcName,
+                                             "indirect, maxdrawcount and stride must all be non-negative."));
+            return false;
+        }
+        const SizeT effectiveStride = stride != 0 ? static_cast<SizeT>(stride) : commandSize;
+        const auto& indirectBuffer =
+            MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::DrawIndirect).GetBoundObject();
+        // A zero maxdrawcount sources nothing, so it cannot run past anything.
+        const SizeT requiredBytes =
+            maxdrawcount == 0 ? 0
+                              : static_cast<SizeT>(indirect) +
+                                    static_cast<SizeT>(maxdrawcount - 1) * effectiveStride + commandSize;
+        if (!indirectBuffer || requiredBytes > indirectBuffer->GetSize()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", funcName,
+                                             "maxdrawcount commands would be sourced from beyond the end of the "
+                                             "buffer bound to GL_DRAW_INDIRECT_BUFFER."));
+            return false;
+        }
+        return true;
+    }
+
     void MultiDrawElementsIndirectCount(GLenum mode, GLenum type, const void* indirect, GLintptr drawcount,
                                         GLsizei maxdrawcount, GLsizei stride) {
+        // Argument validation before the backend-availability check: see DispatchComputeIndirect.
+        // DrawElementsIndirectCommand: count, instanceCount, firstIndex, baseVertex, baseInstance.
+        if (!ValidateIndirectCountDraw(reinterpret_cast<GLintptr>(indirect), drawcount, maxdrawcount, stride,
+                                       5 * sizeof(Uint32), __func__)) {
+            return;
+        }
+        // The only two draw entry points that were missing this. Every backend draw path
+        // dereferences GetProgramForDraw() unconditionally, so "no current program" has to be
+        // stopped here or it is a null dereference rather than the INVALID_OPERATION the spec
+        // asks for - reachable through a bound pipeline that supplies no graphics stage.
+        //
+        // AFTER the argument checks, unlike the sibling draw entry points, and deliberately:
+        // the argument rules here are properties of the call rather than of GL state, and
+        // NegativeApiErrorsTest.IndirectParameterDrawsCheckBothBuffers pins the INVALID_VALUE
+        // they produce for a call made with no program bound. Same precedence decision, and
+        // the same reason, as DispatchComputeIndirect above.
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
         auto multiDrawElementsIndirectCount = MG_Backend::gBackendFunctionsTable.GL.MultiDrawElementsIndirectCount;
         if (!multiDrawElementsIndirectCount) {
             MG_State::pGLContext->RecordError(
@@ -562,6 +694,14 @@ namespace MobileGL::MG_Impl::GLImpl {
 
     void MultiDrawArraysIndirectCount(GLenum mode, const void* indirect, GLintptr drawcount,
                                       GLsizei maxdrawcount, GLsizei stride) {
+        // Argument validation before the backend-availability check: see DispatchComputeIndirect.
+        // DrawArraysIndirectCommand: count, instanceCount, first, baseInstance.
+        if (!ValidateIndirectCountDraw(reinterpret_cast<GLintptr>(indirect), drawcount, maxdrawcount, stride,
+                                       4 * sizeof(Uint32), __func__)) {
+            return;
+        }
+        // See MultiDrawElementsIndirectCount, including why this one goes last.
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
         auto multiDrawArraysIndirectCount = MG_Backend::gBackendFunctionsTable.GL.MultiDrawArraysIndirectCount;
         if (!multiDrawArraysIndirectCount) {
             MG_State::pGLContext->RecordError(
@@ -707,7 +847,7 @@ namespace MobileGL::MG_Impl::GLImpl {
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "Transform feedback is already active."));
             return;
         }
-        const auto& program = MG_State::pGLContext->GetCurrentProgram();
+        const auto& program = MG_State::pGLContext->GetProgramForDraw();
         if (!program || !program->GetLinkStatus() || program->GetTransformFeedbackVaryingCount() == 0) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
@@ -1119,7 +1259,15 @@ namespace MobileGL::MG_Impl::GLImpl {
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName, "instancecount must be non-negative."));
             return;
         }
-        if (!MG_State::pGLContext->ValidateTransformFeedbackName(id)) {
+        // "id is not the name of a transform feedback object" has to mean the same thing here
+        // as it does to glIsTransformFeedback, and the two predicates are not interchangeable:
+        // a name glGenTransformFeedbacks handed out is only reserved until it is first bound,
+        // and only the bind turns it into an object (GL 4.6 core 13.2.1). ValidateTransformFeedbackName
+        // answers the reservation question - the right one for glBindTransformFeedback, which is
+        // what turns a reserved name into an object - so using it here let a generated-but-unbound
+        // name through to the completed-span check below and raised INVALID_OPERATION where the
+        // spec asks for INVALID_VALUE. Name 0 is the default object and always drawable.
+        if (id != 0 && !MG_State::pGLContext->IsTransformFeedbackObject(id)) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue,
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,

MobileGL/MG_Impl/GLImpl/Exporting/Definitions.cpp

@@ -25,12 +25,12 @@
 #define DECLARE_GL_FUNCTION_STUB_HEAD(type, name, ...) MOBILEGL_GL_API type gl##name(__VA_ARGS__) {
 
 #define DECLARE_GL_FUNCTION_STUB_END(type, name, ...)                                                                  \
-    MGLOG_W("Stub function: %s(...)", __FUNCTION__);                                                                   \
+    MGLOG_W_ONCE("Stub function: %s(...)", __FUNCTION__);                                                                   \
     return (type)1;                                                                                                    \
     }
 
 #define DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(type, name, ...)                                                        \
-    MGLOG_W("Stub function: %s(...)", __FUNCTION__);                                                                   \
+    MGLOG_W_ONCE("Stub function: %s(...)", __FUNCTION__);                                                                   \
     }
 
 #define DECLARE_GL_FUNCTION_HEAD(type, name, ...) MOBILEGL_GL_API type gl##name(__VA_ARGS__) {
@@ -320,7 +320,7 @@ DECLARE_GL_FUNCTION_HEAD(void, GetProgramResourceiv, GLuint program, GLenum prog
 DECLARE_GL_FUNCTION_HEAD(GLint, GetProgramResourceLocation, GLuint program, GLenum programInterface, const GLchar* name) DECLARE_GL_FUNCTION_END(GLint, GetProgramResourceLocation, program, programInterface, name)
 DECLARE_GL_FUNCTION_HEAD(void, UseProgramStages, GLuint pipeline, GLbitfield stages, GLuint program) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UseProgramStages, pipeline, stages, program)
 DECLARE_GL_FUNCTION_HEAD(void, ActiveShaderProgram, GLuint pipeline, GLuint program) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ActiveShaderProgram, pipeline, program)
-DECLARE_GL_FUNCTION_STUB_HEAD(GLuint, CreateShaderProgramv, GLenum type, GLsizei count, const GLchar* const* strings) DECLARE_GL_FUNCTION_STUB_END(GLuint, CreateShaderProgramv, type, count, strings)
+DECLARE_GL_FUNCTION_HEAD(GLuint, CreateShaderProgramv, GLenum type, GLsizei count, const GLchar* const* strings) DECLARE_GL_FUNCTION_END(GLuint, CreateShaderProgramv, type, count, strings)
 DECLARE_GL_FUNCTION_HEAD(void, BindProgramPipeline, GLuint pipeline) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindProgramPipeline, pipeline)
 DECLARE_GL_FUNCTION_HEAD(void, DeleteProgramPipelines, GLsizei n, const GLuint* pipelines) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DeleteProgramPipelines, n, pipelines)
 DECLARE_GL_FUNCTION_HEAD(void, GenProgramPipelines, GLsizei n, GLuint* pipelines) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GenProgramPipelines, n, pipelines)
@@ -969,24 +969,24 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, VertexAttribL3dv, GLuint index, const GLdoub
 DECLARE_GL_FUNCTION_STUB_HEAD(void, VertexAttribL4dv, GLuint index, const GLdouble* v) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, VertexAttribL4dv, index, v)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, VertexAttribLPointer, GLuint index, GLint size, GLenum type, GLsizei stride, const void* pointer) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, VertexAttribLPointer, index, size, type, stride, pointer)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetVertexAttribLdv, GLuint index, GLenum pname, GLdouble* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetVertexAttribLdv, index, pname, params)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ViewportArrayv, GLuint first, GLsizei count, const GLfloat* v) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ViewportArrayv, first, count, v)
+DECLARE_GL_FUNCTION_HEAD(void, ViewportArrayv, GLuint first, GLsizei count, const GLfloat* v) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ViewportArrayv, first, count, v)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ViewportIndexedf, GLuint index, GLfloat x, GLfloat y, GLfloat w, GLfloat h) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ViewportIndexedf, index, x, y, w, h)
+DECLARE_GL_FUNCTION_HEAD(void, ViewportIndexedf, GLuint index, GLfloat x, GLfloat y, GLfloat w, GLfloat h) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ViewportIndexedf, index, x, y, w, h)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ViewportIndexedfv, GLuint index, const GLfloat* v) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ViewportIndexedfv, index, v)
+DECLARE_GL_FUNCTION_HEAD(void, ViewportIndexedfv, GLuint index, const GLfloat* v) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ViewportIndexedfv, index, v)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ScissorArrayv, GLuint first, GLsizei count, const GLint* v) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ScissorArrayv, first, count, v)
+DECLARE_GL_FUNCTION_HEAD(void, ScissorArrayv, GLuint first, GLsizei count, const GLint* v) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ScissorArrayv, first, count, v)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ScissorIndexed, GLuint index, GLint left, GLint bottom, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ScissorIndexed, index, left, bottom, width, height)
+DECLARE_GL_FUNCTION_HEAD(void, ScissorIndexed, GLuint index, GLint left, GLint bottom, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ScissorIndexed, index, left, bottom, width, height)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ScissorIndexedv, GLuint index, const GLint* v) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ScissorIndexedv, index, v)
+DECLARE_GL_FUNCTION_HEAD(void, ScissorIndexedv, GLuint index, const GLint* v) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ScissorIndexedv, index, v)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, DepthRangeArrayv, GLuint first, GLsizei count, const GLdouble* v) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DepthRangeArrayv, first, count, v)
+DECLARE_GL_FUNCTION_HEAD(void, DepthRangeArrayv, GLuint first, GLsizei count, const GLdouble* v) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DepthRangeArrayv, first, count, v)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, DepthRangeIndexed, GLuint index, GLdouble n, GLdouble f) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DepthRangeIndexed, index, n, f)
+DECLARE_GL_FUNCTION_HEAD(void, DepthRangeIndexed, GLuint index, GLdouble n, GLdouble f) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DepthRangeIndexed, index, n, f)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetFloati_v, GLenum target, GLuint index, GLfloat* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetFloati_v, target, index, data)
+DECLARE_GL_FUNCTION_HEAD(void, GetFloati_v, GLenum target, GLuint index, GLfloat* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetFloati_v, target, index, data)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetDoublei_v, GLenum target, GLuint index, GLdouble* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetDoublei_v, target, index, data)
+DECLARE_GL_FUNCTION_HEAD(void, GetDoublei_v, GLenum target, GLuint index, GLdouble* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetDoublei_v, target, index, data)
 DECLARE_GL_FUNCTION_HEAD(void, DrawArraysInstancedBaseInstance, GLenum mode, GLint first, GLsizei count, GLsizei instancecount, GLuint baseinstance) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawArraysInstancedBaseInstance, mode, first, count, instancecount, baseinstance)
 DECLARE_GL_FUNCTION_HEAD(void, DrawElementsInstancedBaseInstance, GLenum mode, GLsizei count, GLenum type, const void* indices, GLsizei instancecount, GLuint baseinstance) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawElementsInstancedBaseInstance, mode, count, type, indices, instancecount, baseinstance)
 DECLARE_GL_FUNCTION_HEAD(void, DrawElementsInstancedBaseVertexBaseInstance, GLenum mode, GLsizei count, GLenum type, const void* indices, GLsizei instancecount, GLint basevertex, GLuint baseinstance) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawElementsInstancedBaseVertexBaseInstance, mode, count, type, indices, instancecount, basevertex, baseinstance)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetActiveAtomicCounterBufferiv, GLuint program, GLuint bufferIndex, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetActiveAtomicCounterBufferiv, program, bufferIndex, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, DrawTransformFeedbackInstanced, GLenum mode, GLuint id, GLsizei instancecount) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawTransformFeedbackInstanced, mode, id, instancecount)
 DECLARE_GL_FUNCTION_HEAD(void, DrawTransformFeedbackStreamInstanced, GLenum mode, GLuint id, GLuint stream, GLsizei instancecount) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawTransformFeedbackStreamInstanced, mode, id, stream, instancecount)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ClearBufferData, GLenum target, GLenum internalformat, GLenum format, GLenum type, const void* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ClearBufferData, target, internalformat, format, type, data)
+DECLARE_GL_FUNCTION_HEAD(void, ClearBufferData, GLenum target, GLenum internalformat, GLenum format, GLenum type, const void* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClearBufferData, target, internalformat, format, type, data)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ClearBufferSubData, GLenum target, GLenum internalformat, GLintptr offset, GLsizeiptr size, GLenum format, GLenum type, const void* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ClearBufferSubData, target, internalformat, offset, size, format, type, data)
+DECLARE_GL_FUNCTION_HEAD(void, ClearBufferSubData, GLenum target, GLenum internalformat, GLintptr offset, GLsizeiptr size, GLenum format, GLenum type, const void* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClearBufferSubData, target, internalformat, offset, size, format, type, data)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetInternalformati64v, GLenum target, GLenum internalformat, GLenum pname, GLsizei count, GLint64* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetInternalformati64v, target, internalformat, pname, count, params)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, InvalidateTexSubImage, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, InvalidateTexSubImage, texture, level, xoffset, yoffset, zoffset, width, height, depth)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, InvalidateTexImage, GLuint texture, GLint level) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, InvalidateTexImage, texture, level)
@@ -1003,9 +1003,9 @@ DECLARE_GL_FUNCTION_HEAD(void, ClearTexImage, GLuint texture, GLint level, GLenu
 DECLARE_GL_FUNCTION_HEAD(void, ClearTexSubImage, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, const void* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClearTexSubImage, texture, level, xoffset, yoffset, zoffset, width, height, depth, format, type, data)
 DECLARE_GL_FUNCTION_HEAD(void, BindBuffersBase, GLenum target, GLuint first, GLsizei count, const GLuint* buffers) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindBuffersBase, target, first, count, buffers)
 DECLARE_GL_FUNCTION_HEAD(void, BindBuffersRange, GLenum target, GLuint first, GLsizei count, const GLuint* buffers, const GLintptr* offsets, const GLsizeiptr* sizes) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindBuffersRange, target, first, count, buffers, offsets, sizes)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, BindTextures, GLuint first, GLsizei count, const GLuint* textures) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BindTextures, first, count, textures)
+DECLARE_GL_FUNCTION_HEAD(void, BindTextures, GLuint first, GLsizei count, const GLuint* textures) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindTextures, first, count, textures)
 DECLARE_GL_FUNCTION_HEAD(void, BindSamplers, GLuint first, GLsizei count, const GLuint* samplers) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindSamplers, first, count, samplers)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, BindImageTextures, GLuint first, GLsizei count, const GLuint* textures) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BindImageTextures, first, count, textures)
+DECLARE_GL_FUNCTION_HEAD(void, BindImageTextures, GLuint first, GLsizei count, const GLuint* textures) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindImageTextures, first, count, textures)
 DECLARE_GL_FUNCTION_HEAD(void, BindVertexBuffers, GLuint first, GLsizei count, const GLuint* buffers, const GLintptr* offsets, const GLsizei* strides) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindVertexBuffers, first, count, buffers, offsets, strides)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, ClipControl, GLenum origin, GLenum depth) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ClipControl, origin, depth)
 DECLARE_GL_FUNCTION_HEAD(void, CreateTransformFeedbacks, GLsizei n, GLuint* ids) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CreateTransformFeedbacks, n, ids)
@@ -1061,7 +1061,7 @@ DECLARE_GL_FUNCTION_HEAD(void, TextureSubImage1D, GLuint texture, GLint level, G
 DECLARE_GL_FUNCTION_HEAD(void, TextureSubImage2D, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLenum type, const void* pixels) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureSubImage2D, texture, level, xoffset, yoffset, width, height, format, type, pixels)
 DECLARE_GL_FUNCTION_HEAD(void, TextureSubImage3D, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, const void* pixels) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureSubImage3D, texture, level, xoffset, yoffset, zoffset, width, height, depth, format, type, pixels)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureSubImage1D, GLuint texture, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const void* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedTextureSubImage1D, texture, level, xoffset, width, format, imageSize, data)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureSubImage2D, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const void* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedTextureSubImage2D, texture, level, xoffset, yoffset, width, height, format, imageSize, data)
+DECLARE_GL_FUNCTION_HEAD(void, CompressedTextureSubImage2D, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const void* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CompressedTextureSubImage2D, texture, level, xoffset, yoffset, width, height, format, imageSize, data)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureSubImage3D, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const void* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedTextureSubImage3D, texture, level, xoffset, yoffset, zoffset, width, height, depth, format, imageSize, data)
 DECLARE_GL_FUNCTION_HEAD(void, CopyTextureSubImage1D, GLuint texture, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CopyTextureSubImage1D, texture, level, xoffset, x, y, width)
 DECLARE_GL_FUNCTION_HEAD(void, CopyTextureSubImage2D, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CopyTextureSubImage2D, texture, level, xoffset, yoffset, x, y, width, height)
@@ -1273,7 +1273,7 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, MultiTexCoord4ivARB, GLenum target, const GL
 DECLARE_GL_FUNCTION_STUB_HEAD(void, MultiTexCoord4sARB, GLenum target, GLshort s, GLshort t, GLshort r, GLshort q) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, MultiTexCoord4sARB, target, s, t, r, q)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, MultiTexCoord4svARB, GLenum target, const GLshort* v) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, MultiTexCoord4svARB, target, v)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetQueryObjectivARB, GLuint id, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetQueryObjectivARB, id, pname, params)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, MaxShaderCompilerThreadsARB, GLuint count) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, MaxShaderCompilerThreadsARB, count)
+DECLARE_GL_FUNCTION_HEAD(void, MaxShaderCompilerThreadsARB, GLuint count) DECLARE_GL_FUNCTION_END_NO_RETURN(void, MaxShaderCompilerThreadsARB, count)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PointParameterfARB, GLenum pname, GLfloat param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PointParameterfARB, pname, param)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PointParameterfvARB, GLenum pname, const GLfloat* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PointParameterfvARB, pname, params)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnTexImageARB, GLenum target, GLint level, GLenum format, GLenum type, GLsizei bufSize, void* img) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetnTexImageARB, target, level, format, type, bufSize, img)
@@ -1381,7 +1381,7 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, WindowPos3ivARB, const GLint* v) DECLARE_GL_
 DECLARE_GL_FUNCTION_STUB_HEAD(void, WindowPos3sARB, GLshort x, GLshort y, GLshort z) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, WindowPos3sARB, x, y, z)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, WindowPos3svARB, const GLshort* v) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, WindowPos3svARB, v)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, BlendBarrierKHR, void) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BlendBarrierKHR, )
-DECLARE_GL_FUNCTION_STUB_HEAD(void, MaxShaderCompilerThreadsKHR, GLuint count) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, MaxShaderCompilerThreadsKHR, count)
+DECLARE_GL_FUNCTION_HEAD(void, MaxShaderCompilerThreadsKHR, GLuint count) DECLARE_GL_FUNCTION_END_NO_RETURN(void, MaxShaderCompilerThreadsKHR, count)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, MultiTexCoord1bOES, GLenum texture, GLbyte s) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, MultiTexCoord1bOES, texture, s)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, MultiTexCoord1bvOES, GLenum texture, const GLbyte* coords) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, MultiTexCoord1bvOES, texture, coords)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, MultiTexCoord2bOES, GLenum texture, GLbyte s, GLbyte t) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, MultiTexCoord2bOES, texture, s, t)
@@ -1849,7 +1849,7 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureImage3DEXT, GLuint texture,
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureImage2DEXT, GLuint texture, GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const void* bits) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedTextureImage2DEXT, texture, target, level, internalformat, width, height, border, imageSize, bits)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureImage1DEXT, GLuint texture, GLenum target, GLint level, GLenum internalformat, GLsizei width, GLint border, GLsizei imageSize, const void* bits) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedTextureImage1DEXT, texture, target, level, internalformat, width, border, imageSize, bits)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureSubImage3DEXT, GLuint texture, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const void* bits) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedTextureSubImage3DEXT, texture, target, level, xoffset, yoffset, zoffset, width, height, depth, format, imageSize, bits)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureSubImage2DEXT, GLuint texture, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const void* bits) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedTextureSubImage2DEXT, texture, target, level, xoffset, yoffset, width, height, format, imageSize, bits)
+DECLARE_GL_FUNCTION_HEAD(void, CompressedTextureSubImage2DEXT, GLuint texture, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const void* bits) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CompressedTextureSubImage2D, texture, level, xoffset, yoffset, width, height, format, imageSize, bits)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureSubImage1DEXT, GLuint texture, GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const void* bits) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedTextureSubImage1DEXT, texture, target, level, xoffset, width, format, imageSize, bits)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetCompressedTextureImageEXT, GLuint texture, GLenum target, GLint lod, void* img) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetCompressedTextureImageEXT, texture, target, lod, img)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedMultiTexImage3DEXT, GLenum texunit, GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLsizei imageSize, const void* bits) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CompressedMultiTexImage3DEXT, texunit, target, level, internalformat, width, height, depth, border, imageSize, bits)
@@ -2585,7 +2585,7 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, BindTransformFeedbackNV, GLenum target, GLui
 DECLARE_GL_FUNCTION_STUB_HEAD(void, DeleteTransformFeedbacksNV, GLsizei n, const GLuint* ids) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DeleteTransformFeedbacksNV, n, ids)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GenTransformFeedbacksNV, GLsizei n, GLuint* ids) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GenTransformFeedbacksNV, n, ids)
 MOBILEGL_GL_API GLboolean glIsTransformFeedbackNV(GLuint id) {
-    MGLOG_W("Stub function: %s(...)", __FUNCTION__);
+    MGLOG_W_ONCE("Stub function: %s(...)", __FUNCTION__);
     return GL_FALSE;
 }
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PauseTransformFeedbackNV, void) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PauseTransformFeedbackNV, )
@@ -3181,5 +3181,5 @@ MOBILEGL_GL_API void glVertexAttribDivisorARB(GLuint index, GLuint divisor) {
 }
 
 MOBILEGL_GL_API void glWindowRectanglesEXT(GLenum mode, GLsizei count, const GLint* box) {
-    MGLOG_W("Stub function: %s(...)", __FUNCTION__);
+    MGLOG_W_ONCE("Stub function: %s(...)", __FUNCTION__);
 }

MobileGL/MG_Impl/GLImpl/Framebuffer/GL_Framebuffer.cpp

@@ -106,9 +106,11 @@ namespace MobileGL::MG_Impl::GLImpl {
         // `capabilityTargetIndex` is the row of the cache the attachment actually lives in;
         // kFormatCapabilityTargetCount asks about the format in general. Asking per target matters
         // because a capability recorded for one of them says nothing about the others: DirectGLES
-        // widens three-channel formats to four channels to keep them renderable as *multisample*
+        // decides each target's substitution against that target's own probe, and a buffer texture
-        // storage, and a format that survives only through that substitution is still texture-only
+        // never gets one at all. This is also where the three-channel widening becomes visible to
-        // on every ordinary target.
+        // the application - a GL_RGB8_SNORM colour attachment on a driver with no renderable
+        // three-channel format answers COMPLETE because the backend stores it as GL_RGBA16F and
+        // recorded FramebufferRenderable in CaveatCaps.
         Bool IsColorInternalFormatRenderable(TextureInternalFormat format, SizeT capabilityTargetIndex) {
             const SizeT formatIndex = static_cast<SizeT>(format);
             if (MG_Backend::pActiveBackendObject && formatIndex < MG_Backend::kFormatCapabilityFormatCount) {
@@ -545,7 +547,7 @@ namespace MobileGL::MG_Impl::GLImpl {
                                       GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter) {
         auto blitNamedFramebuffer = MG_Backend::gBackendFunctionsTable.GL.BlitNamedFramebuffer;
         if (!blitNamedFramebuffer) {
-            MGLOG_E("glBlitNamedFramebuffer skipped: backend does not implement explicit framebuffer blit.");
+            MGLOG_E_ONCE("glBlitNamedFramebuffer skipped: backend does not implement explicit framebuffer blit.");
             return;
         }
         blitNamedFramebuffer(readFramebuffer, drawFramebuffer, srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1,
@@ -556,7 +558,7 @@ namespace MobileGL::MG_Impl::GLImpl {
                                          GLenum buffer, GLint drawbuffer, const GLfloat* value) {
         auto clearNamedFramebufferfv = MG_Backend::gBackendFunctionsTable.GL.ClearNamedFramebufferfv;
         if (!clearNamedFramebufferfv) {
-            MGLOG_E("glClearNamedFramebufferfv skipped: backend does not implement explicit framebuffer clear.");
+            MGLOG_E_ONCE("glClearNamedFramebufferfv skipped: backend does not implement explicit framebuffer clear.");
             return;
         }
         clearNamedFramebufferfv(framebuffer, buffer, drawbuffer, value);
@@ -566,7 +568,7 @@ namespace MobileGL::MG_Impl::GLImpl {
                                          GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil) {
         auto clearNamedFramebufferfi = MG_Backend::gBackendFunctionsTable.GL.ClearNamedFramebufferfi;
         if (!clearNamedFramebufferfi) {
-            MGLOG_E("glClearNamedFramebufferfi skipped: backend does not implement explicit framebuffer clear.");
+            MGLOG_E_ONCE("glClearNamedFramebufferfi skipped: backend does not implement explicit framebuffer clear.");
             return;
         }
         clearNamedFramebufferfi(framebuffer, buffer, drawbuffer, depth, stencil);
@@ -576,7 +578,7 @@ namespace MobileGL::MG_Impl::GLImpl {
                                          GLenum buffer, GLint drawbuffer, const GLint* value) {
         auto clearNamedFramebufferiv = MG_Backend::gBackendFunctionsTable.GL.ClearNamedFramebufferiv;
         if (!clearNamedFramebufferiv) {
-            MGLOG_E("glClearNamedFramebufferiv skipped: backend does not implement explicit framebuffer clear.");
+            MGLOG_E_ONCE("glClearNamedFramebufferiv skipped: backend does not implement explicit framebuffer clear.");
             return;
         }
         clearNamedFramebufferiv(framebuffer, buffer, drawbuffer, value);
@@ -586,7 +588,7 @@ namespace MobileGL::MG_Impl::GLImpl {
                                           GLenum buffer, GLint drawbuffer, const GLuint* value) {
         auto clearNamedFramebufferuiv = MG_Backend::gBackendFunctionsTable.GL.ClearNamedFramebufferuiv;
         if (!clearNamedFramebufferuiv) {
-            MGLOG_E("glClearNamedFramebufferuiv skipped: backend does not implement explicit framebuffer clear.");
+            MGLOG_E_ONCE("glClearNamedFramebufferuiv skipped: backend does not implement explicit framebuffer clear.");
             return;
         }
         clearNamedFramebufferuiv(framebuffer, buffer, drawbuffer, value);
@@ -618,6 +620,34 @@ namespace MobileGL::MG_Impl::GLImpl {
         return std::max(MG_Backend::pActiveBackendObject->GetDynamicParameters().MaxSamples, 1);
     }
 
+    // GL_MAX_SAMPLES is the ceiling over all formats; an integer format has its own, lower
+    // one (GL_MAX_INTEGER_SAMPLES) and GL 4.6 core 9.2.4 makes exceeding it INVALID_OPERATION.
+    // The multisample TEXTURE path already resolves the limit per format
+    // (GL_Texture.cpp, GetMaxTextureSamplesForFormat); renderbuffers only ever compared
+    // against GL_MAX_SAMPLES, so on a driver where the two differ - Adreno reports
+    // GL_MAX_SAMPLES 4 and GL_MAX_INTEGER_SAMPLES 1 - an integer renderbuffer accepted a
+    // sample count the format cannot deliver, and said GL_NO_ERROR about it.
+    Int GetMaxRenderbufferSamplesForFormat_State(TextureInternalFormat format) {
+        if (MG_Backend::pActiveBackendObject == nullptr) {
+            return std::numeric_limits<Int>::max();
+        }
+        const auto& dynamicParameters = MG_Backend::pActiveBackendObject->GetDynamicParameters();
+
+        GLenum normalizedInternalFormat = MG_Util::ConvertTextureInternalFormatToGLEnum(format);
+        GLenum normalizedFormat = GL_RGBA;
+        GLenum normalizedType = GL_UNSIGNED_BYTE;
+        MG_Util::TextureFormatProcessor::NormalizePixelFormat(normalizedInternalFormat,
+                                                              PixelFormatNormalizeOptionBit::None,
+                                                              &normalizedInternalFormat, &normalizedFormat,
+                                                              &normalizedType);
+        const Bool isIntegerFormat = normalizedFormat == GL_RED_INTEGER || normalizedFormat == GL_RG_INTEGER ||
+                                     normalizedFormat == GL_RGB_INTEGER || normalizedFormat == GL_RGBA_INTEGER;
+        if (!isIntegerFormat) {
+            return GetMaxRenderbufferSamples_State();
+        }
+        return std::max(dynamicParameters.MaxIntegerSamples, 1);
+    }
+
     Bool ValidateRenderbufferStorageSize_State(GLsizei width, GLsizei height, const char* caller) {
         if (width < 0 || height < 0) {
             MG_State::pGLContext->RecordError(
@@ -639,7 +669,7 @@ namespace MobileGL::MG_Impl::GLImpl {
         return true;
     }
 
-    Bool ValidateRenderbufferStorageSamples_State(GLsizei samples, const char* caller) {
+    Bool ValidateRenderbufferStorageSamples_State(GLsizei samples, TextureInternalFormat format, const char* caller) {
         if (samples < 0) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue,
@@ -647,9 +677,10 @@ namespace MobileGL::MG_Impl::GLImpl {
             return false;
         }
 
-        const Int maxSamples = GetMaxRenderbufferSamples_State();
+        // TODO: Resolve the remaining per-internalformat renderbuffer sample limits once
+        // glGetInternalformativ is backed; integer formats are handled below.
+        const Int maxSamples = GetMaxRenderbufferSamplesForFormat_State(format);
         if (samples > maxSamples) {
-            // TODO: Use per-internalformat renderbuffer sample limits once glGetInternalformativ is backed.
             // GL 4.6 core 9.2.4 makes asking for more samples than the format supports
             // INVALID_OPERATION, not INVALID_VALUE - the count is well formed, this format just
             // cannot deliver it. Only a negative count is INVALID_VALUE.
@@ -657,7 +688,7 @@ namespace MobileGL::MG_Impl::GLImpl {
                 ErrorCode::InvalidOperation,
                 MakeUnique<GenericErrorInfo>(
                     "MG_Impl/GLImpl", caller,
-                    std::format("Sample count {} exceeds GL_MAX_SAMPLES ({}).", samples, maxSamples)));
+                    std::format("Sample count {} exceeds this format's sample limit ({}).", samples, maxSamples)));
             return false;
         }
         return true;
@@ -682,7 +713,7 @@ namespace MobileGL::MG_Impl::GLImpl {
         TextureInternalFormat format = MG_Util::ConvertGLEnumToTextureInternalFormat(internalformat);
         if (!TextureImpl::ValidateTextureInternalFormat(format)) return;
 
-        if (!ValidateRenderbufferStorageSamples_State(samples, kCaller)) return;
+        if (!ValidateRenderbufferStorageSamples_State(samples, format, kCaller)) return;
         if (!ValidateRenderbufferStorageSize_State(width, height, kCaller)) return;
 
         renderbufferObject->AllocateStorage({width, height});
@@ -929,7 +960,8 @@ namespace MobileGL::MG_Impl::GLImpl {
 
         TextureInternalFormat format = MG_Util::ConvertGLEnumToTextureInternalFormat(internalformat);
         if (!TextureImpl::ValidateTextureInternalFormat(format)) return;
-        if (!ValidateRenderbufferStorageSamples_State(samples, "NamedRenderbufferStorageMultisample_State")) return;
+        if (!ValidateRenderbufferStorageSamples_State(samples, format, "NamedRenderbufferStorageMultisample_State"))
+            return;
         if (!ValidateRenderbufferStorageSize_State(width, height, "NamedRenderbufferStorageMultisample_State")) return;
 
         renderbufferObject->AllocateStorage({width, height});
@@ -1069,11 +1101,36 @@ namespace MobileGL::MG_Impl::GLImpl {
         case TextureTarget::Texture2DMultisampleArray:
             textureUploadTarget = TextureUploadTarget::Texture2DMultisampleArray;
             break;
+        case TextureTarget::Texture1DArray:
+            textureUploadTarget = TextureUploadTarget::Texture1DArray;
+            break;
+        case TextureTarget::TextureCubeMapArray:
+            textureUploadTarget = TextureUploadTarget::CubeMapArray;
+            break;
         default:
-            RecordUnsupportedFramebufferTextureAttachmentError(
+            MG_State::pGLContext->RecordError(
-                __func__, "FramebufferTextureLayer requires a 3D, 2D array or 2D multisample array texture.");
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "FramebufferTextureLayer requires a 3D, array, 2D multisample "
+                                             "array, or cube map array texture."));
             return;
         }
+        // The same backend question the DSA twin asks. GL 4.6 core 9.2.8 makes the two entry points
+        // equivalent, so they have to decline in the same places - leaving this one ungated is what
+        // let an unrepresentable attachment reach the renderer, and it also refused cube map arrays
+        // that GL requires it to accept.
+        {
+            const auto& layerLimits = MG_Backend::pActiveBackendObject
+                                          ? MG_Backend::pActiveBackendObject->GetDynamicParameters()
+                                          : MG_Backend::DynamicBackendParameters{};
+            const TextureTarget layeredTarget = textureObject->GetTarget();
+            if ((layer != 0 || layeredTarget == TextureTarget::TextureCubeMapArray) &&
+                !layerLimits.SupportsPerLayerFramebufferAttachment(layeredTarget)) {
+                RecordUnsupportedFramebufferTextureAttachmentError(
+                    __func__, "This backend does not resolve a framebuffer attachment's layer onto its image.");
+                return;
+            }
+        }
         AttachFramebufferTextureLayer(__func__, target, attachment, texture, level, layer, textureUploadTarget);
     }
 
@@ -1433,11 +1490,19 @@ namespace MobileGL::MG_Impl::GLImpl {
         // so a slice lands outside the image and the renderer asserts on the clear. Letting it
         // through there would only move the failure downstream, so it is declined instead - layer
         // zero always works, being the plain first-slice attachment.
-        const Bool backsLayeredAttachment = limits.SupportsPerLayerFramebufferAttachment;
+        // ...and it is a DIFFERENT question per target: a 2D/2D-multisample array layer is a Vulkan
-        // A cube map array additionally has no image shape at all in VkTextureManager, so on that
+        // array layer, a 3D layer is a z slice, and a cube map array needs a cube-compatible image
-        // backend it cannot be an attachment whatever the layer is.
+        // before it has any layer to name. Ask the backend about this texture's target rather than
-        const Bool isCubeMapArray = textureObject->GetTarget() == TextureTarget::TextureCubeMapArray;
+        // guessing from one blanket flag.
-        if ((layer != 0 && !backsLayeredAttachment) || (isCubeMapArray && !backsLayeredAttachment)) {
+        const TextureTarget layeredTextureTarget = textureObject->GetTarget();
+        const Bool backsThisTargetsLayers = limits.SupportsPerLayerFramebufferAttachment(layeredTextureTarget);
+        // Layer zero of a 3D or array texture is the plain first-slice attachment every backend can
+        // already express, so it stays legal even where per-layer selection is not backed. A cube map
+        // array has no such fallback: layer zero is still one face of one cube inside a
+        // cube-compatible image, so it needs the same support layer 5 does.
+        const Bool needsPerLayerSupport =
+            layer != 0 || layeredTextureTarget == TextureTarget::TextureCubeMapArray;
+        if (needsPerLayerSupport && !backsThisTargetsLayers) {
             RecordUnsupportedFramebufferTextureAttachmentError(
                 __func__, "This backend does not resolve a framebuffer attachment's layer onto its image.");
             return;

MobileGL/MG_Impl/GLImpl/Getter/GL_Getter.cpp

@@ -16,6 +16,7 @@
 #include <MG_State/GLState/ErrorState/ErrorInfo.h>
 #include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
 #include <MG_Util/Converters/GLToMG/BufferEnumConverter.h>
+#include <MG_Util/Converters/GLToMG/RenderStateEnumConverter.h>
 #include <MG_Util/Converters/MGToGL/FramebufferEnumConverter.h>
 #include <MG_Util/Converters/MGToGL/ErrorCodeConverter.h>
 #include <MG_Util/Converters/MGToGL/TextureEnumConverter.h>
@@ -23,9 +24,15 @@
 #include <MG_Util/Converters/MGToGL/RenderStateEnumConverter.h>
 #include <MG_State/GLState/FramebufferState/FramebufferObject.h>
 #include <MG_Util/Texture/TextureFormatProcessor.h>
+#include <MG_Util/Async/ShaderCompilePool.h>
 #include <MG_Backend/BackendObjects.h>
 
 namespace MobileGL::MG_Impl::GLImpl {
+    // Declared rather than #included from GL_RenderState.h on purpose: that header also declares
+    // a free function named BlendEquation, which would hide the ::MobileGL::BlendEquation enum
+    // this file's blend-state queries name unqualified.
+    GLboolean IsEnabledi(GLenum target, GLuint index);
+
     namespace {
         enum class IndexedBufferQueryKind {
             Binding,
@@ -50,6 +57,23 @@ namespace MobileGL::MG_Impl::GLImpl {
         constexpr GLint kFrontendMaxTessControlAtomicCounters = 0;
         constexpr GLint kFrontendMaxTessEvaluationAtomicCounters = 0;
         constexpr GLint kFrontendMaxVertexAtomicCounters = 0;
+        // Zero counters means zero buffers to hold them. These have to be ANSWERED rather than
+        // left to the default INVALID_ENUM: a well-behaved application queries the limit exactly
+        // to find out that the stage cannot do this, and an error instead both leaves its output
+        // untouched (so it reads uninitialised memory and may conclude the opposite) and leaves a
+        // GL error pending that surfaces at whatever unrelated call checks next.
+        constexpr GLint kFrontendMaxGeometryAtomicCounterBuffers = 0;
+        constexpr GLint kFrontendMaxTessControlAtomicCounterBuffers = 0;
+        constexpr GLint kFrontendMaxTessEvaluationAtomicCounterBuffers = 0;
+        constexpr GLint kFrontendMaxVertexAtomicCounterBuffers = 0;
+        // One atomic counter is a uint, and a buffer never has to hold more counters than the
+        // combined limit the frontend advertises. GL 4.6 table 23.63 floors this at 32 bytes.
+        constexpr GLint kFrontendMaxAtomicCounterBufferSize =
+            kFrontendMaxCombinedAtomicCounters * static_cast<GLint>(sizeof(GLuint));
+        // KHR_debug minima (GL 4.6 table 23.66); the debug entry points are stubs, but the
+        // limits they advertise still have to be legal.
+        constexpr GLint kFrontendMaxDebugGroupStackDepth = 64;
+        constexpr GLint kFrontendMaxDebugLoggedMessages = 1;
         constexpr GLint kFrontendMaxVertexUniformComponents = 4096;
         constexpr GLint kFrontendMaxVertexUniformVectors = 128;
         constexpr GLint kFrontendMaxVertexUniformBlocks = 14;
@@ -165,6 +189,30 @@ namespace MobileGL::MG_Impl::GLImpl {
             return frontendCount;
         }
 
+        // A per-stage or combined BLOCK count is an amount of indexed binding points an
+        // application will occupy, and GL 4.6 table 23.64 orders the two accordingly:
+        // MAX_UNIFORM_BUFFER_BINDINGS >= MAX_COMBINED_UNIFORM_BLOCKS >= every per-stage count,
+        // and the same for the shader-storage family. The two families are answered from
+        // unrelated places here - frontend constants, backend dynamic parameters, and a few
+        // hard-coded TODOs - so nothing kept them ordered, and a backend that reports Vulkan
+        // descriptor-indexing counts advertised 256 compute uniform blocks over 36 binding
+        // points. KHR-GL44.multi_bind.dispatch_bind_buffers_base reads the block count and binds
+        // that many buffers in ONE glBindBuffersBase, which is then INVALID_OPERATION before it
+        // binds anything. Clamping is the only direction available: the binding count is the
+        // capacity of the state layer's indexed-binding array, not a number we may inflate.
+        GLint ClampBlockCountToBindingPoints(GLint blockCount, BufferTarget bufferTarget) {
+            const GLint bindingPoints = static_cast<GLint>(GetIndexedBufferQueryPointCount(bufferTarget));
+            return std::min(std::max(blockCount, 0), bindingPoints);
+        }
+
+        GLint ClampUniformBlockCount(GLint blockCount) {
+            return ClampBlockCountToBindingPoints(blockCount, BufferTarget::Uniform);
+        }
+
+        GLint ClampStorageBlockCount(GLint blockCount) {
+            return ClampBlockCountToBindingPoints(blockCount, BufferTarget::ShaderStorage);
+        }
+
         bool TryDecodeDrawBufferQuery(GLenum pname, SizeT& drawBufferIndex) {
             if (pname == GL_DRAW_BUFFER) {
                 drawBufferIndex = 0;
@@ -263,6 +311,104 @@ namespace MobileGL::MG_Impl::GLImpl {
             return true;
         }
 
+        // GL_TEXTURE_BINDING_* is per-texture-unit state: glGetIntegerv answers for the
+        // active unit, glGetIntegeri_v answers for unit `index`. Both need the same
+        // pname -> target decode, so it lives here instead of being spelled out twice.
+        bool TryDecodeTextureUnitBindingPname(GLenum pname, TextureTarget& outTarget) {
+            switch (pname) {
+            case GL_TEXTURE_BINDING_1D: outTarget = TextureTarget::Texture1D; return true;
+            case GL_TEXTURE_BINDING_1D_ARRAY: outTarget = TextureTarget::Texture1DArray; return true;
+            case GL_TEXTURE_BINDING_2D: outTarget = TextureTarget::Texture2D; return true;
+            case GL_TEXTURE_BINDING_2D_ARRAY: outTarget = TextureTarget::Texture2DArray; return true;
+            case GL_TEXTURE_BINDING_2D_MULTISAMPLE: outTarget = TextureTarget::Texture2DMultisample; return true;
+            case GL_TEXTURE_BINDING_2D_MULTISAMPLE_ARRAY:
+                outTarget = TextureTarget::Texture2DMultisampleArray;
+                return true;
+            case GL_TEXTURE_BINDING_3D: outTarget = TextureTarget::Texture3D; return true;
+            case GL_TEXTURE_BINDING_BUFFER: outTarget = TextureTarget::TextureBuffer; return true;
+            case GL_TEXTURE_BINDING_CUBE_MAP: outTarget = TextureTarget::TextureCubeMap; return true;
+            case GL_TEXTURE_BINDING_CUBE_MAP_ARRAY: outTarget = TextureTarget::TextureCubeMapArray; return true;
+            case GL_TEXTURE_BINDING_RECTANGLE: outTarget = TextureTarget::TextureRectangle; return true;
+            default: return false;
+            }
+        }
+
+        GLint QueryTextureBindingOnUnit(Int unit, TextureTarget target) {
+            auto& textureUnit = MG_State::pGLContext->GetTextureUnitObject(unit);
+            const auto& obj = textureUnit.GetBindingSlot(target).GetBoundObject();
+            return obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
+        }
+
+        GLint QuerySamplerBindingOnUnit(Int unit) {
+            const auto& textureUnit = MG_State::pGLContext->GetTextureUnitObject(unit);
+            const auto& sampler = textureUnit.GetSamplerObject();
+            return sampler ? static_cast<GLint>(sampler->GetExternalIndex()) : 0;
+        }
+
+        // The ARB_viewport_array indexed rectangles. Each of these is genuinely per-viewport
+        // frontend state (RenderStateParameters::Viewports / ScissorBoxes / DepthRanges), so the
+        // indexed getters must read the indexed storage - the generic path at the bottom of
+        // GetIntegeri_v is a raw backend passthrough that has no case for them and returned
+        // zeros, and routing them to the NON-indexed getter (what this used to do) answered every
+        // index with viewport 0's value, which is what
+        // KHR-GL43.viewport_array.{viewport,scissor,depth_range}_api caught.
+        Bool IsIndexedViewportQuery(GLenum target) {
+            return target == GL_VIEWPORT || target == GL_SCISSOR_BOX || target == GL_DEPTH_RANGE;
+        }
+
+        // Component count of an indexed viewport-array query, so every width of getter writes the
+        // caller's whole buffer instead of just element 0 (GL 4.6 core 22.1).
+        GLsizei IndexedViewportQueryComponents(GLenum target) {
+            return target == GL_DEPTH_RANGE ? 2 : 4;
+        }
+
+        // ARB_viewport_array: `index` selects a viewport and MAX_VIEWPORTS bounds it. The bound is
+        // the frontend's own state width, which is also exactly what GL_MAX_VIEWPORTS reports -
+        // taking it from the backend caps instead would let a device limit of 1 (a Vulkan device
+        // without the multiViewport feature) make index 1 illegal even though the state exists.
+        Bool ValidateViewportQueryIndex(GLuint index, const char* caller) {
+            if (index < RenderStateParameters::MAX_VIEWPORTS) return true;
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", caller, "Viewport index is out of range."));
+            return false;
+        }
+
+        // The indexed viewport/scissor/depth-range state as floats, which is the widest lossless
+        // shape MobileGL stores (the viewport really is float state; the scissor box is integral
+        // and well inside float's exact range, and every depth range is in [0, 1]). Every indexed
+        // getter width funnels through this so they can never disagree with each other.
+        void ReadIndexedViewportStateFloat(GLenum target, GLuint index, GLfloat* out) {
+            switch (target) {
+            case GL_VIEWPORT: {
+                const FloatVec4& viewport = MG_State::pGLContext->GetViewportIndexed(index);
+                out[0] = viewport.x();
+                out[1] = viewport.y();
+                out[2] = viewport.z();
+                out[3] = viewport.w();
+                return;
+            }
+            case GL_SCISSOR_BOX: {
+                const IntVec4& box = MG_State::pGLContext->GetScissorBoxIndexed(index);
+                out[0] = static_cast<GLfloat>(box.x());
+                out[1] = static_cast<GLfloat>(box.y());
+                out[2] = static_cast<GLfloat>(box.z());
+                out[3] = static_cast<GLfloat>(box.w());
+                return;
+            }
+            case GL_DEPTH_RANGE: {
+                const FloatVec2& range = MG_State::pGLContext->GetDepthRangeIndexed(index);
+                out[0] = range.x();
+                out[1] = range.y();
+                return;
+            }
+            default:
+                MOBILEGL_ASSERT(false, "ReadIndexedViewportStateFloat: unexpected target 0x%x",
+                                static_cast<Uint32>(target));
+                return;
+            }
+        }
+
         void CopyIntsToBooleans(const GLint* src, SizeT count, GLboolean* dst) {
             for (SizeT i = 0; i < count; ++i) {
                 dst[i] = src[i] ? GL_TRUE : GL_FALSE;
@@ -287,7 +433,7 @@ namespace MobileGL::MG_Impl::GLImpl {
 
         MGLOG_D("glGetString, name: %s", MG_Util::ConvertGLEnumToString(name).c_str());
         if (!activeBackendObject) {
-            MGLOG_E("activeBackendObject is not initialized!");
+            MGLOG_E_ONCE("activeBackendObject is not initialized!");
             return (GLubyte*)"Unknown";
         }
 
@@ -346,7 +492,7 @@ namespace MobileGL::MG_Impl::GLImpl {
 
         const auto& activeBackendObject = MG_Backend::pActiveBackendObject;
         if (!activeBackendObject) {
-            MGLOG_E("activeBackendObject is not initialized!");
+            MGLOG_E_ONCE("activeBackendObject is not initialized!");
             return (GLubyte*)"Unknown";
         }
         const auto& rendererInfo = activeBackendObject->GetRendererInfo();
@@ -533,6 +679,17 @@ namespace MobileGL::MG_Impl::GLImpl {
             params[1] = dynamicParameters.ViewportBoundsRangeMax;
             return;
         }
+        // Viewport 0's rectangle, verbatim. Falling through to the integer width below would
+        // round the fractional rectangle a glViewportIndexedf(0, ...) is allowed to set, and
+        // glGetFloatv(GL_VIEWPORT) is a lossless query of float state.
+        case GL_VIEWPORT: {
+            const FloatVec4& viewport = MG_State::pGLContext->GetViewportIndexed(0);
+            params[0] = viewport.x();
+            params[1] = viewport.y();
+            params[2] = viewport.z();
+            params[3] = viewport.w();
+            return;
+        }
         case GL_MIN_FRAGMENT_INTERPOLATION_OFFSET:
         case GL_MAX_FRAGMENT_INTERPOLATION_OFFSET:
         case GL_FRAGMENT_INTERPOLATION_OFFSET_BITS: {
@@ -659,10 +816,14 @@ namespace MobileGL::MG_Impl::GLImpl {
                     *data = 0;
                     return;
                 }
+                // GL 4.6 core table 23.4/23.5: *_BUFFER_SIZE reports the size glBindBufferRange
+                // was ASKED for, verbatim. It is not clamped to the buffer's storage, and it does
+                // not follow the buffer when a later glBufferData resizes it - a range may legally
+                // name bytes the buffer does not have yet. Clamping it here answered 0 for the
+                // common conformance shape of binding a range on a buffer that has no storage
+                // yet (KHR-GL43.shader_storage_buffer_object.basic-binding).
                 const Range1D range = bindingPoint.GetRange();
-                const auto start = std::min(range.start, bufferObject->GetSize());
+                *data = static_cast<GLint>(range.end - range.start);
-                const auto end = std::min(range.end, bufferObject->GetSize());
-                *data = static_cast<GLint>(end - start);
                 return;
             }
             default:
@@ -670,7 +831,54 @@ namespace MobileGL::MG_Impl::GLImpl {
             }
         }
 
+        // Per-texture-unit bindings: GL 4.6 core table 23.19 makes every GL_TEXTURE_BINDING_*
+        // and GL_SAMPLER_BINDING indexed by texture unit. Without this they fell through to
+        // the raw backend passthrough at the bottom, which knows nothing about the
+        // frontend's binding state.
+        if (TextureTarget textureBindingTarget = TextureTarget::Unknown;
+            TryDecodeTextureUnitBindingPname(target, textureBindingTarget) || target == GL_SAMPLER_BINDING) {
+            GLint maxUnits = 0;
+            GetIntegerv(GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS, &maxUnits);
+            maxUnits = std::min<GLint>(maxUnits, MG_State::GLState::TextureState::MAX_TEXTURE_IMAGE_UNITS);
+            if (index >= static_cast<GLuint>(std::max(maxUnits, 0))) {
+                *data = 0;
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "Texture unit index is out of range."));
+                return;
+            }
+            *data = target == GL_SAMPLER_BINDING
+                ? QuerySamplerBindingOnUnit(static_cast<Int>(index))
+                : QueryTextureBindingOnUnit(static_cast<Int>(index), textureBindingTarget);
+            return;
+        }
+
+        // GL 4.6 core 22.1: an indexed query answers EVERY indexed state, and GL_SCISSOR_TEST is
+        // indexed by viewport just like GL_BLEND is by draw buffer. Without this the integer
+        // width fell through to the backend passthrough and answered GL_INVALID_ENUM, which is
+        // the sticky error KHR-GL43.viewport_array.queries trips over at its next error check.
+        if (MG_Util::ConvertGLEnumToCapabilityInput(target) != CapabilityInput::Unknown) {
+            *data = IsEnabledi(target, index);
+            return;
+        }
+
         switch (target) {
+        // ARB_viewport_array queries the indexed rectangles through glGetIntegeri_v as well
+        // (gl4cMultiBindTests and the viewport_array group both do).
+        case GL_VIEWPORT:
+        case GL_SCISSOR_BOX:
+        case GL_DEPTH_RANGE: {
+            if (!ValidateViewportQueryIndex(index, __func__)) return;
+            GLfloat values[4] = {};
+            ReadIndexedViewportStateFloat(target, index, values);
+            const GLsizei components = IndexedViewportQueryComponents(target);
+            for (GLsizei i = 0; i < components; ++i) {
+                // Round, not truncate: glGetIntegerv on floating-point state rounds to nearest
+                // (GL 4.6 core 22.2), so a 255.875-wide viewport reads back as 256 and not 255.
+                data[i] = static_cast<GLint>(std::lround(values[i]));
+            }
+            return;
+        }
         // The vertex buffer binding points of the vertex array object that is bound. Indexed by
         // binding point, not by attribute (GL 4.6 core 10.3.1).
         case GL_VERTEX_BINDING_BUFFER:
@@ -782,6 +990,54 @@ namespace MobileGL::MG_Impl::GLImpl {
         getIntegeri(target, index, data);
     }
 
+    // GL_ARB_viewport_array's typed indexed getters. They were no-op stubs, which left the
+    // caller's output buffer holding whatever was on the stack. The multi-component indexed
+    // rectangles are answered from the frontend's own viewport/scissor/depth-range state, via
+    // the non-indexed getter of the matching type - GL_DEPTH_RANGE is float state, so putting
+    // it through the integer query would round it to 0/1. Everything else MobileGL answers
+    // indexed is scalar integer-domain state, where converting the integer query is exact.
+    void GetFloati_v(GLenum target, GLuint index, GLfloat* data) {
+        if (!data) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "data pointer cannot be null"));
+            return;
+        }
+        if (IsIndexedViewportQuery(target)) {
+            if (!ValidateViewportQueryIndex(index, __func__)) return;
+            // Verbatim, NOT via the integer width: the viewport is float state and
+            // KHR-GL43.viewport_array.viewport_api compares the read-back with ==, so a
+            // glViewportIndexedf(i, 0.125f, ...) has to come back as 0.125f exactly.
+            ReadIndexedViewportStateFloat(target, index, data);
+            return;
+        }
+        GLint ints[4] = {};
+        GetIntegeri_v(target, index, ints);
+        data[0] = static_cast<GLfloat>(ints[0]);
+    }
+
+    void GetDoublei_v(GLenum target, GLuint index, GLdouble* data) {
+        if (!data) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "data pointer cannot be null"));
+            return;
+        }
+        if (IsIndexedViewportQuery(target)) {
+            if (!ValidateViewportQueryIndex(index, __func__)) return;
+            GLfloat values[4] = {};
+            ReadIndexedViewportStateFloat(target, index, values);
+            const GLsizei components = IndexedViewportQueryComponents(target);
+            for (GLsizei i = 0; i < components; ++i) {
+                data[i] = static_cast<GLdouble>(values[i]);
+            }
+            return;
+        }
+        GLint ints[4] = {};
+        GetIntegeri_v(target, index, ints);
+        data[0] = static_cast<GLdouble>(ints[0]);
+    }
+
     void GetInteger64i_v(GLenum target, GLuint index, GLint64* data) {
         if (!data) {
             MG_State::pGLContext->RecordError(
@@ -818,9 +1074,8 @@ namespace MobileGL::MG_Impl::GLImpl {
                     *data = 0;
                     return;
                 }
-                const auto start = std::min(range.start, bufferObject->GetSize());
+                // Verbatim, unclamped - see the GetIntegeri_v arm.
-                const auto end = std::min(range.end, bufferObject->GetSize());
+                *data = static_cast<GLint64>(range.end - range.start);
-                *data = static_cast<GLint64>(end - start);
                 return;
             }
             default:
@@ -828,15 +1083,35 @@ namespace MobileGL::MG_Impl::GLImpl {
             }
         }
 
-        auto getInteger64i = MG_Backend::gBackendFunctionsTable.GL.GetInteger64i_v;
+        // The one indexed pname whose value genuinely needs 64 bits: a vertex buffer binding
-        if (!getInteger64i) {
+        // offset is an intptr, so taking the 32-bit route below would truncate it.
-            *data = 0;
+        if (target == GL_VERTEX_BINDING_OFFSET) {
+            if (index >= VertexArrayImpl::GetMaxVertexAttribBindings()) {
                 MG_State::pGLContext->RecordError(
-                ErrorCode::InvalidOperation,
+                    ErrorCode::InvalidValue,
-                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "Backend does not support indexed integer queries."));
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                                 "Vertex buffer binding index is out of range."));
                 return;
             }
-        getInteger64i(target, index, data);
+            const auto& vao = MG_State::pGLContext->GetBoundVertexArray();
+            *data = vao ? static_cast<GLint64>(vao->GetBindingPoint(index).Offset) : 0;
+            return;
+        }
+
+        // Everything else is 32-bit indexed state that the glGetIntegeri_v pname table already
+        // owns, and GL 4.6 core 22.1 says every indexed query answers every indexed pname.
+        // Handing the leftovers straight to the backend instead made glGetInteger64i_v disagree
+        // with glGetIntegeri_v on the very same pname - GL_MAX_COMPUTE_WORK_GROUP_COUNT read
+        // back 0 while the 32-bit view said 65535 (KHR-GL43.compute_shader.max), because a
+        // frontend-only value simply is not in the driver's table.
+        GLint values[4] = {};
+        GetIntegeri_v(target, index, values);
+        // The viewport-array rectangles are the only multi-component indexed state here; every
+        // other pname is scalar, so widening element 0 alone would silently truncate them.
+        const GLsizei components = IsIndexedViewportQuery(target) ? IndexedViewportQueryComponents(target) : 1;
+        for (GLsizei i = 0; i < components; ++i) {
+            data[i] = static_cast<GLint64>(values[i]);
+        }
     }
 
     void GetInteger64v(GLenum pname, GLint64* params) {
@@ -958,6 +1233,13 @@ namespace MobileGL::MG_Impl::GLImpl {
             return;
         }
 
+        // Per-texture-unit bindings: the non-indexed query reports the active unit.
+        if (TextureTarget textureBindingTarget = TextureTarget::Unknown;
+            TryDecodeTextureUnitBindingPname(pname, textureBindingTarget)) {
+            *params = QueryTextureBindingOnUnit(MG_State::pGLContext->GetActiveTextureUnit(), textureBindingTarget);
+            return;
+        }
+
         switch (pname) {
         case GL_ACTIVE_TEXTURE:
             *params = MG_State::pGLContext->GetActiveTextureUnit() + GL_TEXTURE0;
@@ -1069,12 +1351,32 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
             return;
         }
+        case GL_MAX_SHADER_COMPILER_THREADS_KHR:
+            // GL_KHR_parallel_shader_compile (GL_MAX_SHADER_COMPILER_THREADS_ARB is the same
+            // 0x91B0). The number of threads MobileGL's compile pool would actually use, so
+            // an application sizing its own submission batches gets a real answer.
+            //
+            // Zero when asynchronous compilation is off, which is the honest reply and the
+            // one the extension defines for an implementation with no compiler threads: the
+            // extension string is withdrawn in that configuration too, so a conforming
+            // application never reaches this query, and one that asks anyway is told there
+            // are none rather than being handed a thread count nothing will use.
+            *params = MG_Util::Async::AsyncShaderCompileEnabled()
+                          ? static_cast<GLint>(MG_Util::Async::ShaderCompilePool::Get().GetThreadCount())
+                          : 0;
+            return;
         case GL_MAX_DEBUG_GROUP_STACK_DEPTH:
-            *params = 0; // debug-group entrypoints are stubbed
+            // KHR_debug floors this at 64 even when the group entry points are stubs: the
+            // limit describes how deep glPushDebugGroup may nest, and 0 is not a legal answer.
+            *params = kFrontendMaxDebugGroupStackDepth;
             return;
         case GL_MAX_DEBUG_MESSAGE_LENGTH:
             *params = 1024; // debug-message entrypoints are stubbed, but KHR_debug requires a valid limit
             return;
+        case GL_MAX_DEBUG_LOGGED_MESSAGES:
+            // Size of the message log ring; KHR_debug requires at least 1.
+            *params = kFrontendMaxDebugLoggedMessages;
+            return;
         case GL_DEBUG_GROUP_STACK_DEPTH:
             *params = 0; // debug-group entrypoints are stubbed
             return;
@@ -1219,7 +1521,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = kFrontendMaxCombinedAtomicCounters;
             return;
         case GL_MAX_COMBINED_UNIFORM_BLOCKS:
-            *params = kFrontendMaxCombinedUniformBlocks;
+            *params = ClampUniformBlockCount(kFrontendMaxCombinedUniformBlocks);
             return;
         case GL_MAX_DUAL_SOURCE_DRAW_BUFFERS:
             *params = 1; // TODO
@@ -1234,7 +1536,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = kFrontendMaxFragmentAtomicCounters;
             return;
         case GL_MAX_FRAGMENT_SHADER_STORAGE_BLOCKS:
-            *params = 16; // TODO
+            *params = ClampStorageBlockCount(16); // TODO
             return;
         case GL_MAX_FRAGMENT_INPUT_COMPONENTS:
             *params = kFrontendMaxFragmentInputComponents;
@@ -1251,13 +1553,16 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = kFrontendMaxFragmentUniformVectors;
             return;
         case GL_MAX_FRAGMENT_UNIFORM_BLOCKS:
-            *params = kFrontendMaxFragmentUniformBlocks;
+            *params = ClampUniformBlockCount(kFrontendMaxFragmentUniformBlocks);
             return;
         case GL_MAX_GEOMETRY_ATOMIC_COUNTERS:
             *params = kFrontendMaxGeometryAtomicCounters;
             return;
+        case GL_MAX_GEOMETRY_ATOMIC_COUNTER_BUFFERS:
+            *params = kFrontendMaxGeometryAtomicCounterBuffers;
+            return;
         case GL_MAX_GEOMETRY_SHADER_STORAGE_BLOCKS:
-            *params = 16; // TODO
+            *params = ClampStorageBlockCount(16); // TODO
             return;
         case GL_MAX_GEOMETRY_INPUT_COMPONENTS:
             *params = kFrontendMaxGeometryInputComponents;
@@ -1280,7 +1585,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = kFrontendMaxGeometryTotalOutputComponents;
             return;
         case GL_MAX_GEOMETRY_UNIFORM_BLOCKS:
-            *params = kFrontendMaxGeometryUniformBlocks;
+            *params = ClampUniformBlockCount(kFrontendMaxGeometryUniformBlocks);
             return;
         case GL_MAX_GEOMETRY_UNIFORM_COMPONENTS:
             *params = kFrontendMaxGeometryUniformComponents;
@@ -1312,9 +1617,15 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_MAX_TESS_CONTROL_ATOMIC_COUNTERS:
             *params = kFrontendMaxTessControlAtomicCounters;
             return;
+        case GL_MAX_TESS_CONTROL_ATOMIC_COUNTER_BUFFERS:
+            *params = kFrontendMaxTessControlAtomicCounterBuffers;
+            return;
         case GL_MAX_TESS_EVALUATION_ATOMIC_COUNTERS:
             *params = kFrontendMaxTessEvaluationAtomicCounters;
             return;
+        case GL_MAX_TESS_EVALUATION_ATOMIC_COUNTER_BUFFERS:
+            *params = kFrontendMaxTessEvaluationAtomicCounterBuffers;
+            return;
         case GL_MAX_TESS_CONTROL_IMAGE_UNIFORMS:
             *params = 0;
             return;
@@ -1322,10 +1633,10 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = 0;
             return;
         case GL_MAX_TESS_CONTROL_SHADER_STORAGE_BLOCKS:
-            *params = 16; // TODO
+            *params = ClampStorageBlockCount(16); // TODO
             return;
         case GL_MAX_TESS_EVALUATION_SHADER_STORAGE_BLOCKS:
-            *params = 16; // TODO
+            *params = ClampStorageBlockCount(16); // TODO
             return;
         case GL_MAX_TEXTURE_LOD_BIAS:
             *params = 15; // TODO
@@ -1342,13 +1653,16 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_MAX_VERTEX_ATOMIC_COUNTERS:
             *params = kFrontendMaxVertexAtomicCounters;
             return;
+        case GL_MAX_VERTEX_ATOMIC_COUNTER_BUFFERS:
+            *params = kFrontendMaxVertexAtomicCounterBuffers;
+            return;
         case GL_MAX_VERTEX_IMAGE_UNIFORMS:
             *params = MG_Backend::pActiveBackendObject
                           ? MG_Backend::pActiveBackendObject->GetDynamicParameters().MaxVertexImageUniforms
                           : MG_Backend::DynamicBackendParameters{}.MaxVertexImageUniforms;
             return;
         case GL_MAX_VERTEX_SHADER_STORAGE_BLOCKS:
-            *params = 16; // TODO
+            *params = ClampStorageBlockCount(16); // TODO
             return;
         case GL_MAX_VERTEX_UNIFORM_COMPONENTS:
             *params = kFrontendMaxVertexUniformComponents;
@@ -1360,7 +1674,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = kFrontendMaxVertexOutputComponents;
             return;
         case GL_MAX_VERTEX_UNIFORM_BLOCKS:
-            *params = kFrontendMaxVertexUniformBlocks;
+            *params = ClampUniformBlockCount(kFrontendMaxVertexUniformBlocks);
             return;
         case GL_NUM_COMPRESSED_TEXTURE_FORMATS:
             *params = 0; // compressed texture upload entrypoints are still unimplemented
@@ -1516,13 +1830,9 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_SAMPLE_MASK_VALUE:
             *params = static_cast<GLint>(MG_State::pGLContext->GetSampleMaskValue());
             return;
-        case GL_SAMPLER_BINDING: {
+        case GL_SAMPLER_BINDING:
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
+            *params = QuerySamplerBindingOnUnit(MG_State::pGLContext->GetActiveTextureUnit());
-            const auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& sampler = tu.GetSamplerObject();
-            *params = sampler ? static_cast<GLint>(sampler->GetExternalIndex()) : 0;
             return;
-        }
         case GL_SAMPLES:
             *params = ResolveDrawFramebufferSampleCount();
             return;
@@ -1618,87 +1928,6 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_STEREO:
             *params = 0; // stereo surfaces are not exposed
             return;
-        case GL_TEXTURE_BINDING_1D: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::Texture1D);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
-        case GL_TEXTURE_BINDING_1D_ARRAY: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::Texture1DArray);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
-        case GL_TEXTURE_BINDING_2D: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::Texture2D);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            MGLOG_D("Get GL_TEXTURE_BINDING_2D: %d", *params);
-            return;
-        }
-        case GL_TEXTURE_BINDING_2D_ARRAY: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::Texture2DArray);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
-        case GL_TEXTURE_BINDING_2D_MULTISAMPLE: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::Texture2DMultisample);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
-        case GL_TEXTURE_BINDING_2D_MULTISAMPLE_ARRAY: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::Texture2DMultisampleArray);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
-        case GL_TEXTURE_BINDING_3D: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::Texture3D);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
-        case GL_TEXTURE_BINDING_BUFFER: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::TextureBuffer);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
-        case GL_TEXTURE_BINDING_CUBE_MAP: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::TextureCubeMap);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
-        case GL_TEXTURE_BINDING_RECTANGLE: {
-            Int unit = MG_State::pGLContext->GetActiveTextureUnit();
-            auto& tu = MG_State::pGLContext->GetTextureUnitObject(unit);
-            const auto& slot = tu.GetBindingSlot(TextureTarget::TextureRectangle);
-            const auto& obj = slot.GetBoundObject();
-            *params = obj ? static_cast<GLint>(obj->GetExternalIndex()) : 0;
-            return;
-        }
         case GL_TEXTURE_COMPRESSION_HINT:
             *params = static_cast<GLint>(MG_State::pGLContext->GetHint(pname));
             return;
@@ -1816,7 +2045,7 @@ namespace MobileGL::MG_Impl::GLImpl {
 
         const auto& activeBackendObject = MG_Backend::pActiveBackendObject;
         if (!activeBackendObject) {
-            MGLOG_E("activeBackendObject is not initialized!");
+            MGLOG_E_ONCE("activeBackendObject is not initialized!");
             return;
         }
         const auto& rendererInfo = activeBackendObject->GetRendererInfo();
@@ -1845,13 +2074,13 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = dynamicParameters.SubgroupQuadOperationsInAllStages ? GL_TRUE : GL_FALSE;
             break;
         case GL_MAX_COMPUTE_SHADER_STORAGE_BLOCKS:
-            *params = dynamicParameters.MaxComputeShaderStorageBlocks;
+            *params = ClampStorageBlockCount(dynamicParameters.MaxComputeShaderStorageBlocks);
             break;
         case GL_MAX_COMBINED_SHADER_STORAGE_BLOCKS:
-            *params = dynamicParameters.MaxCombinedShaderStorageBlocks;
+            *params = ClampStorageBlockCount(dynamicParameters.MaxCombinedShaderStorageBlocks);
             break;
         case GL_MAX_COMPUTE_UNIFORM_BLOCKS:
-            *params = dynamicParameters.MaxComputeUniformBlocks;
+            *params = ClampUniformBlockCount(dynamicParameters.MaxComputeUniformBlocks);
             break;
         case GL_MAX_COMPUTE_TEXTURE_IMAGE_UNITS:
             *params = dynamicParameters.MaxComputeTextureImageUnits;
@@ -1971,6 +2200,26 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS:
             *params = static_cast<GLint>(GetIndexedBufferQueryPointCount(BufferTarget::ShaderStorage));
             break;
+        case GL_MAX_SHADER_STORAGE_BLOCK_SIZE:
+            // 64-bit state (see GetInteger64v); the 32-bit query saturates, per the GL
+            // state-query conversion rules.
+            *params = static_cast<GLint>(std::min<Uint64>(dynamicParameters.MaxShaderStorageBlockSize,
+                                                          static_cast<Uint64>(INT32_MAX)));
+            break;
+        case GL_MAX_ATOMIC_COUNTER_BUFFER_BINDINGS:
+            *params = static_cast<GLint>(GetIndexedBufferQueryPointCount(BufferTarget::AtomicCounter));
+            break;
+        case GL_MAX_ATOMIC_COUNTER_BUFFER_SIZE:
+            // The conformance suite splits this evenly across every advertised binding point and
+            // binds all of them in one glBindBuffersRange
+            // (KHR-GL44.multi_bind.functional_bind_buffers_range), so the pair has to divide:
+            // 32 bytes over 36 binding points is a zero-sized range, which BindBufferRange
+            // rejects with INVALID_VALUE before it binds anything. Floor the advertised size at
+            // one counter per binding point.
+            *params = std::max<GLint>(
+                kFrontendMaxAtomicCounterBufferSize,
+                static_cast<GLint>(GetIndexedBufferQueryPointCount(BufferTarget::AtomicCounter) * sizeof(GLuint)));
+            break;
         case GL_MAX_TEXTURE_BUFFER_SIZE:
             *params = dynamicParameters.MaxTextureBufferSize;
             break;
@@ -2034,7 +2283,15 @@ namespace MobileGL::MG_Impl::GLImpl {
             params[1] = dynamicParameters.MaxViewportHeight;
             break;
         case GL_MAX_VIEWPORTS:
-            *params = dynamicParameters.MaxViewports;
+            // The frontend's own state width, not the backend's device limit. GL 4.3 core
+            // requires MAX_VIEWPORTS >= 16 and every indexed viewport entry point validates
+            // against RenderStateParameters::MAX_VIEWPORTS, so reporting anything else would
+            // either advertise viewports the state cannot hold or reject indices it can. A
+            // Vulkan device without the multiViewport feature reports maxViewports == 1, which
+            // limits what can be RASTERIZED to more than one rectangle (see the multiViewport
+            // gate in VulkanRenderer), not what the GL state can hold; caps.MaxViewports keeps
+            // carrying that device number for exactly that decision.
+            *params = static_cast<GLint>(RenderStateParameters::MAX_VIEWPORTS);
             break;
         case GL_MINOR_VERSION:
             *params = rendererInfo.RendererGLInfo.TargetGLVersion.Minor;
@@ -2090,7 +2347,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             *params = static_cast<GLint>(std::lround(dynamicParameters.MaxTextureMaxAnisotropy));
             break;
         default:
-            MGLOG_E("glGetIntegerv: Invalid enum %s (0x%X)", MG_Util::ConvertGLEnumToString(pname).c_str(), pname);
+            MGLOG_D("glGetIntegerv: Invalid enum %s (0x%X)", MG_Util::ConvertGLEnumToString(pname).c_str(), pname);
             MG_State::pGLContext->RecordError(ErrorCode::InvalidEnum,
                                               MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "GetIntegerv",
                                                                            std::format("Invalid enum: 0x{:X}", pname)));

MobileGL/MG_Impl/GLImpl/Getter/GL_Getter.h

@@ -19,6 +19,8 @@ namespace MobileGL::MG_Impl::GLImpl {
     void GetIntegerv(GLenum pname, GLint* params);
     void GetInteger64v(GLenum pname, GLint64* params);
     void GetIntegeri_v(GLenum target, GLuint index, GLint* data);
+    void GetFloati_v(GLenum target, GLuint index, GLfloat* data);
+    void GetDoublei_v(GLenum target, GLuint index, GLdouble* data);
     void GetInteger64i_v(GLenum target, GLuint index, GLint64* data);
     GLenum GetError();
     GLenum GetGraphicsResetStatus();

MobileGL/MG_Impl/GLImpl/Program/GL_Program.h

@@ -42,6 +42,10 @@ namespace MobileGL::MG_Impl::GLImpl {
     GLboolean IsProgram(GLuint program);
     GLboolean IsShader(GLuint shader);
     void LinkProgram(GLuint program);
+    // GL_KHR_parallel_shader_compile / GL_ARB_parallel_shader_compile. Both names are the
+    // same entry point; see MaxShaderCompilerThreadsKHR_State for the semantics of count.
+    void MaxShaderCompilerThreadsKHR(GLuint count);
+    void MaxShaderCompilerThreadsARB(GLuint count);
     void ShaderSource(GLuint shader, GLsizei count, const GLchar* const* string, const GLint* length);
     void UseProgram(GLuint program);
     void Uniform1f(GLint location, GLfloat v0);
@@ -174,6 +178,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     void GetUniformdv(GLuint program, GLint location, GLdouble* params);
     void ValidateProgram(GLuint program);
     void ProgramParameteri(GLuint program, GLenum pname, GLint value);
+    GLuint CreateShaderProgramv(GLenum type, GLsizei count, const GLchar* const* strings);
     void GetProgramBinary(GLuint program, GLsizei bufSize, GLsizei* length, GLenum* binaryFormat, void* binary);
     void ProgramBinary(GLuint program, GLenum binaryFormat, const void* binary, GLsizei length);
     void TransformFeedbackVaryings(GLuint program, GLsizei count, const GLchar* const* varyings, GLenum bufferMode);

MobileGL/MG_Impl/GLImpl/Program/GL_ProgramPipeline.cpp

@@ -19,16 +19,26 @@ namespace MobileGL::MG_Impl::GLImpl {
                 code, MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", function, Move(message)));
         }
 
-        // A pipeline name only names an object once it has been bound or created; querying a
+        // GL 4.6 core 7.4 asks only that the name came from GenProgramPipelines and has not been
-        // reserved-but-unmaterialised name is INVALID_OPERATION (GL 4.6 core 7.4).
+        // deleted - so a name that was reserved and never bound is legal here, and the command
+        // MATERIALIZES it rather than rejecting it.
+        //
+        // Requiring a bound object instead is what broke every separable-program conformance case
+        // across three families: the CTS reserves a name, calls glUseProgramStages three times and
+        // only then binds, which is the order the spec's own example uses. Each of those calls
+        // failed with INVALID_OPERATION, so the stage programs were never recorded - the pipeline
+        // stayed empty, GetProgramForDraw flattened nothing and the draw painted nothing, and the
+        // rejected calls' error was left in the queue for the harness to find. One cause, both
+        // symptoms.
         const SharedPtr<MG_State::GLState::ProgramPipelineObject>* TryGetPipeline(GLuint pipeline,
                                                                                  const char* function) {
-            if (!MG_State::pGLContext->IsProgramPipelineObject(pipeline)) {
+            const auto& object = MG_State::pGLContext->MaterializeProgramPipelineObject(pipeline);
+            if (!object) {
                 RecordPipelineError(ErrorCode::InvalidOperation, function,
                                     std::format("Program pipeline {} does not exist.", pipeline));
                 return nullptr;
             }
-            return &MG_State::pGLContext->GetProgramPipelineObject(pipeline);
+            return &object;
         }
 
         Bool ValidatePipelineCount(GLsizei n, const char* function) {

MobileGL/MG_Impl/GLImpl/Program/ProgramInterface.cpp

@@ -0,0 +1,918 @@
+// MobileGL - MobileGL/MG_Impl/GLImpl/Program/ProgramInterface.cpp
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#include "ProgramInterface.h"
+
+#include <MG_State/GLState/ProgramState/ProgramObject.h>
+#include <MG_Util/ShaderTranspiler/Types.h>
+
+#include <cstring>
+
+namespace MobileGL::MG_Impl::GLImpl::ProgramInterface {
+    namespace {
+        // glslang folds atomic counters into synthesized blocks named
+        // "<getAtomicCounterBlockName()>_<binding>" (ParseContextBase.cpp), one per GL
+        // atomic-counter binding point. That block IS the GL_ATOMIC_COUNTER_BUFFER resource
+        // and its trailing number IS GL_BUFFER_BINDING; its members stay GL_UNIFORMs.
+        constexpr const char* kAtomicCounterBlockPrefix = "gl_AtomicCounterBlock";
+
+        enum class BlockKind {
+            Uniform,       // a real GL uniform block
+            GlobalUbo,     // the synthesized MGL_GLOBAL_UBO: GL sees its members as default-block
+            AtomicCounter, // gl_AtomicCounterBlock_<binding>
+            Storage,       // a shader storage block
+        };
+
+        // One row of any interface. Fields a given interface does not have keep the
+        // spec-mandated "not applicable" value, so a prop read never has to special-case
+        // the interface a second time.
+        struct Resource {
+            String name;
+            GLenum type = GL_NONE;
+            GLint arraySize = 1;
+            GLint location = -1;
+            GLint locationIndex = -1;
+            GLint blockIndex = -1;
+            GLint offset = -1;
+            GLint arrayStride = -1;
+            GLint matrixStride = -1;
+            GLint isRowMajor = 0;
+            GLint atomicCounterBufferIndex = -1;
+            GLint topLevelArraySize = 0;
+            GLint topLevelArrayStride = 0;
+            GLint bufferBinding = 0;
+            GLint bufferDataSize = 0;
+            GLint isPerPatch = 0;
+            GLint xfbBufferIndex = 0;
+            Uint32 stages = 0; // EShLanguageMask
+            Vector<GLuint> activeVariables;
+        };
+
+        using ResourceList = Vector<Resource>;
+
+        struct Model {
+            ResourceList uniforms;
+            ResourceList uniformBlocks;
+            ResourceList atomicCounterBuffers;
+            ResourceList bufferVariables;
+            ResourceList storageBlocks;
+            ResourceList programInputs;
+            ResourceList programOutputs;
+            ResourceList xfbVaryings;
+            Bool valid = false;
+        };
+
+        const ResourceList& EmptyList() {
+            static const ResourceList empty;
+            return empty;
+        }
+
+        // ---- name spelling (cluster 6) -------------------------------------------------
+
+        Bool EndsWithZeroSubscript(const String& name) {
+            return name.length() >= 3 && name.compare(name.length() - 3, 3, "[0]") == 0;
+        }
+
+        // The enumerated spelling of an array resource is "name[0]". glslang already applies
+        // that to uniforms and buffer variables (EShReflectionBasicArraySuffix), but never to
+        // stage inputs/outputs, so those get it here.
+        String WithArraySuffix(const String& name, const glslang::TType* type) {
+            if (type == nullptr || !type->isArray() || EndsWithZeroSubscript(name)) return name;
+            return name + "[0]";
+        }
+
+        // GL_ARRAY_SIZE: element count for a sized array, 0 for a runtime-sized one
+        // (a shader storage block's unsized trailing member), 1 for a non-array.
+        GLint ArraySizeOf(const glslang::TType* type, GLint reflectedSize) {
+            if (type != nullptr && type->isArray()) {
+                if (!type->isSizedArray()) return 0;
+                return type->getOuterArraySize();
+            }
+            return reflectedSize < 1 ? 1 : reflectedSize;
+        }
+
+        // Two spellings name the same resource when they are equal, or differ only by the
+        // "[0]" the enumeration appends to an array.
+        Bool NamesMatch(const String& resourceName, const String& query) {
+            if (resourceName == query) return true;
+            if (EndsWithZeroSubscript(resourceName) &&
+                resourceName.compare(0, resourceName.length() - 3, query) == 0) {
+                return true;
+            }
+            return EndsWithZeroSubscript(query) && query.compare(0, query.length() - 3, resourceName) == 0;
+        }
+
+        // Splits "base[k]" into ("base", k). GL 4.6 §7.3.1.1 requires the subscript to be a
+        // decimal integer with no white space and no leading zeros, which is exactly what
+        // separates array-names' "a[1]" (resolves) from "a[01]", "a[0 + 0]" and "a[ 0]" (do
+        // not). Returns false when there is no trailing subscript at all; sets `malformed`
+        // when there is one but it is not a strict decimal.
+        Bool SplitTrailingSubscript(const String& name, String& outBase, Uint& outElement, Bool& outMalformed) {
+            outMalformed = false;
+            if (name.empty() || name.back() != ']') return false;
+            const SizeT bracket = name.rfind('[');
+            if (bracket == String::npos) return false;
+            const SizeT first = bracket + 1;
+            const SizeT last = name.length() - 1; // one past the digits
+            if (first >= last) {
+                outMalformed = true;
+                return false;
+            }
+            // No leading zeros: "0" is the only spelling that may start with '0'.
+            if (name[first] == '0' && last - first > 1) {
+                outMalformed = true;
+                return false;
+            }
+            Uint element = 0;
+            for (SizeT i = first; i < last; ++i) {
+                if (name[i] < '0' || name[i] > '9') {
+                    outMalformed = true;
+                    return false;
+                }
+                element = element * 10 + static_cast<Uint>(name[i] - '0');
+                if (element > 0x0FFFFFFFu) {
+                    outMalformed = true;
+                    return false;
+                }
+            }
+            outBase = name.substr(0, bracket);
+            outElement = element;
+            return true;
+        }
+
+        // ---- block classification ------------------------------------------------------
+
+        Bool IsAtomicCounterBlockName(const String& name) {
+            return name.compare(0, std::strlen(kAtomicCounterBlockPrefix), kAtomicCounterBlockPrefix) == 0;
+        }
+
+        // "gl_AtomicCounterBlock_5" -> 5. The suffix is the GL binding the counters were
+        // declared with, which glslang does NOT keep in the block's own layout qualifier
+        // (that one is remapped to a plain buffer binding).
+        GLint AtomicCounterBlockBinding(const String& name) {
+            const SizeT underscore = name.rfind('_');
+            if (underscore == String::npos || underscore + 1 >= name.length()) return 0;
+            GLint binding = 0;
+            for (SizeT i = underscore + 1; i < name.length(); ++i) {
+                if (name[i] < '0' || name[i] > '9') return 0;
+                binding = binding * 10 + (name[i] - '0');
+            }
+            return binding;
+        }
+
+        // Element index of an arrayed block instance ("TrickyBuffer[1]" -> 1).
+        GLint BlockArrayElement(const String& name) {
+            String base;
+            Uint element = 0;
+            Bool malformed = false;
+            if (!SplitTrailingSubscript(name, base, element, malformed)) return 0;
+            return static_cast<GLint>(element);
+        }
+
+        BlockKind ClassifyBlock(const glslang::TObjectReflection& block) {
+            if (std::strstr(block.name.c_str(), MG_Util::ShaderTranspiler::GLOBAL_UBO_NAME) != nullptr) {
+                return BlockKind::GlobalUbo;
+            }
+            if (IsAtomicCounterBlockName(block.name)) return BlockKind::AtomicCounter;
+            const glslang::TType* type = block.getType();
+            if (type != nullptr && type->getQualifier().storage == glslang::EvqBuffer) return BlockKind::Storage;
+            return BlockKind::Uniform;
+        }
+
+        // std140/std430 column stride, the same vec4-rounded rule ProgramObject applies to
+        // uniform matrices. 0 for a non-matrix.
+        GLint MatrixStrideOf(const glslang::TType* type) {
+            if (type == nullptr || !type->isMatrix()) return 0;
+            const bool rowMajor = type->getQualifier().layoutMatrix == glslang::ElmRowMajor;
+            const int strideVectorComponents = rowMajor ? type->getMatrixCols() : type->getMatrixRows();
+            constexpr int scalarSize = 4;
+            const int vectorAlignment = (strideVectorComponents <= 1)    ? scalarSize
+                                        : (strideVectorComponents == 2) ? 2 * scalarSize
+                                                                        : 4 * scalarSize;
+            return (vectorAlignment + 15) & ~15;
+        }
+
+        GLint IsRowMajorOf(const glslang::TType* type) {
+            if (type == nullptr || !type->isMatrix()) return 0;
+            return type->getQualifier().layoutMatrix == glslang::ElmRowMajor ? 1 : 0;
+        }
+
+        GLint MappedLocation(Int rawLocation) {
+            // glslang parks "no location" at layoutLocationEnd; GL spells it -1.
+            if (rawLocation < 0 || rawLocation >= static_cast<Int>(glslang::TQualifier::layoutLocationEnd)) return -1;
+            return rawLocation;
+        }
+
+        // ---- model construction --------------------------------------------------------
+
+        // GL_REFERENCED_BY_*_SHADER for an ARRAYED block instance, refined per element.
+        //
+        // glslang records a block reference by walking up to the base symbol and calling
+        // addBlockName with the whole ARRAY type, which ORs the referencing stage into every
+        // element at once - it has not resolved the subscript yet at that point. So reading
+        // "e[0].b" marks both TrickyBlock[0] and TrickyBlock[1] as referenced by the fragment
+        // stage (KHR-GL43.program_interface_query.uniform-block-types).
+        //
+        // The MEMBER masks are exact: EShReflectionAllBlockVariables enumerates every member of
+        // every element with the stage mask suppressed, and only the dereference chain actually
+        // walked turns a bit on - and that chain carries the subscript. So the union of a block
+        // instance's members is the reference set of that instance.
+        //
+        // Applied ONLY to arrayed instances, because for a scalar block glslang is already exact.
+        // Note the union is used even when it is empty: an array element nobody dereferenced has
+        // no member bits and is genuinely referenced by nobody, which is the whole point - falling
+        // back to the block's own mask there would restore the over-approximation.
+        Vector<Uint32> BuildBlockStagesFromMembers(const glslang::TProgram& reflection, Int blockCount) {
+            auto& mutableReflection = const_cast<glslang::TProgram&>(reflection);
+            Vector<Uint32> stagesByBlock(static_cast<SizeT>(blockCount < 0 ? 0 : blockCount), 0u);
+            const Int uniformCount = mutableReflection.getNumUniformVariables();
+            for (Int index = 0; index < uniformCount; ++index) {
+                const auto& uniform = mutableReflection.getUniform(index);
+                const Int owner = uniform.index;
+                if (owner < 0 || owner >= blockCount) continue;
+                stagesByBlock[static_cast<SizeT>(owner)] |= static_cast<Uint32>(uniform.stages);
+            }
+            return stagesByBlock;
+        }
+
+        // UNIFORM blocks only, and that scope is load-bearing rather than cautious. The member
+        // names glslang produces for a uniform block array carry the subscript
+        // ("TrickyBlock[0].b", via EShReflectionStrictArraySuffix), so each element's members are
+        // distinct entries and the bits land on the right one. A SHADER STORAGE block array does
+        // NOT get that treatment - its buffer variables reflect under one subscript-free spelling
+        // shared by every element - so a union over them credits element 0 and starves the rest.
+        // KHR-GL43.program_interface_query.ssb-types is the case that says so: it reads ss[0] and
+        // ss[1] and requires both to report the fragment stage, which only glslang's own
+        // (deliberately over-approximating) block mask gets right. Storage and atomic-counter
+        // blocks therefore keep that mask untouched.
+        Uint32 UniformBlockStages(const glslang::TObjectReflection& block, const Vector<Uint32>& stagesFromMembers,
+                                  Int tIndex) {
+            String arrayBase;
+            Uint element = 0;
+            Bool malformed = false;
+            if (!SplitTrailingSubscript(block.name, arrayBase, element, malformed) || malformed) {
+                return static_cast<Uint32>(block.stages);
+            }
+            if (tIndex < 0 || tIndex >= static_cast<Int>(stagesFromMembers.size())) {
+                return static_cast<Uint32>(block.stages);
+            }
+            return stagesFromMembers[static_cast<SizeT>(tIndex)];
+        }
+
+        void BuildBlocks(ProgramObject& program, const glslang::TProgram& reflection, Model& model,
+                         Vector<BlockKind>& blockKind, Vector<Int>& blockInterfaceIndex) {
+            const Int blockCount = const_cast<glslang::TProgram&>(reflection).getNumUniformBlocks();
+            blockKind.assign(blockCount, BlockKind::Uniform);
+            blockInterfaceIndex.assign(blockCount, -1);
+            const Vector<Uint32> stagesFromMembers = BuildBlockStagesFromMembers(reflection, blockCount);
+
+            for (Int tIndex = 0; tIndex < blockCount; ++tIndex) {
+                const auto& block = const_cast<glslang::TProgram&>(reflection).getUniformBlock(tIndex);
+                const BlockKind kind = ClassifyBlock(block);
+                blockKind[tIndex] = kind;
+                if (kind == BlockKind::AtomicCounter) {
+                    Resource resource;
+                    // GL_ATOMIC_COUNTER_BUFFER resources have no name (and GetProgramResource
+                    // Index/Name reject the interface outright, which is why this stays empty).
+                    resource.bufferBinding = AtomicCounterBlockBinding(block.name);
+                    resource.bufferDataSize = block.size;
+                    resource.stages = static_cast<Uint32>(block.stages);
+                    blockInterfaceIndex[tIndex] = static_cast<Int>(model.atomicCounterBuffers.size());
+                    model.atomicCounterBuffers.push_back(Move(resource));
+                } else if (kind == BlockKind::Storage) {
+                    Resource resource;
+                    resource.name = block.name;
+                    // glslang reports the DECLARED binding for every instance of an arrayed
+                    // block; GL gives element k the binding base + k. That is only the initial
+                    // value: GL_BUFFER_BINDING must report the CURRENT binding, so a later
+                    // glShaderStorageBlockBinding wins over the declaration (GL 4.6 §7.6.2 -
+                    // exactly the same rule GL_UNIFORM_BLOCK follows through
+                    // GetUniformBlockBinding below).
+                    const GLint declared = block.getBinding();
+                    resource.bufferBinding = declared < 0 ? 0 : declared + BlockArrayElement(block.name);
+                    const Int rebound = program.GetShaderStorageBlockBindingOverride(block.name);
+                    if (rebound >= 0) resource.bufferBinding = static_cast<GLint>(rebound);
+                    resource.bufferDataSize = block.size;
+                    resource.stages = static_cast<Uint32>(block.stages);
+                    blockInterfaceIndex[tIndex] = static_cast<Int>(model.storageBlocks.size());
+                    model.storageBlocks.push_back(Move(resource));
+                }
+            }
+
+            // GL_UNIFORM_BLOCK keeps the index space glUniformBlockBinding and
+            // glGetActiveUniformBlockiv already use, so an index handed out here is usable
+            // with them (which is exactly what the CTS does).
+            const Int glBlockCount = program.GetActiveUniformBlocksCount();
+            for (Int glIndex = 0; glIndex < glBlockCount; ++glIndex) {
+                Resource resource;
+                resource.name = program.GetUniformBlockName(glIndex);
+                resource.bufferBinding = static_cast<GLint>(program.GetUniformBlockBinding(glIndex));
+                resource.bufferDataSize = static_cast<GLint>(program.GetUBOSizeAt(glIndex));
+                const Int tIndex = program.TProgramBlockIndex(static_cast<Uint>(glIndex));
+                if (tIndex >= 0 && tIndex < blockCount) {
+                    resource.stages = UniformBlockStages(const_cast<glslang::TProgram&>(reflection).getUniformBlock(tIndex),
+                                                      stagesFromMembers, tIndex);
+                }
+                model.uniformBlocks.push_back(Move(resource));
+            }
+        }
+
+        void BuildUniformsAndBufferVariables(ProgramObject& program, const glslang::TProgram& reflection, Model& model,
+                                             const Vector<BlockKind>& blockKind,
+                                             const Vector<Int>& blockInterfaceIndex) {
+            const Uint uniformCount = program.GetUniformCount();
+            for (Uint glIndex = 0; glIndex < uniformCount; ++glIndex) {
+                const Int tIndex = program.TProgramUniformIndex(glIndex);
+                const auto& refl = const_cast<glslang::TProgram&>(reflection).getUniform(tIndex);
+                const glslang::TType* type = refl.getType();
+                const Int owner = refl.index;
+                const BlockKind kind = (owner >= 0 && owner < static_cast<Int>(blockKind.size()))
+                                           ? blockKind[owner]
+                                           : BlockKind::GlobalUbo;
+
+                Resource resource;
+                resource.name = refl.name;
+                resource.type = static_cast<GLenum>(refl.glDefineType);
+                resource.arraySize = ArraySizeOf(type, refl.size);
+                resource.stages = static_cast<Uint32>(refl.stages);
+
+                if (kind == BlockKind::Storage) {
+                    resource.blockIndex = blockInterfaceIndex[owner];
+                    resource.offset = refl.offset;
+                    resource.arrayStride = refl.arrayStride;
+                    resource.matrixStride = MatrixStrideOf(type);
+                    resource.isRowMajor = IsRowMajorOf(type);
+                    // GL requires 1 for a member that is not inside a top-level array (and for
+                    // the top-level array itself); glslang leaves 0/-1 there.
+                    resource.topLevelArraySize = refl.topLevelArraySize > 0 ? refl.topLevelArraySize : 1;
+                    resource.topLevelArrayStride = refl.topLevelArrayStride;
+                    model.bufferVariables.push_back(Move(resource));
+                    continue;
+                }
+
+                if (kind == BlockKind::AtomicCounter) {
+                    // An atomic counter is a default-block uniform with no location and no
+                    // owning uniform block; what it does have is a buffer to point at.
+                    resource.type = GL_UNSIGNED_INT_ATOMIC_COUNTER;
+                    resource.blockIndex = -1;
+                    resource.offset = refl.offset;
+                    resource.arrayStride = refl.arrayStride;
+                    resource.matrixStride = 0;
+                    resource.atomicCounterBufferIndex = blockInterfaceIndex[owner];
+                    resource.location = -1;
+                } else {
+                    resource.blockIndex = program.GetActiveUniformBlockIndex(glIndex);
+                    resource.offset = program.GetActiveUniformOffset(glIndex);
+                    resource.arrayStride = program.GetActiveUniformArrayStride(glIndex);
+                    resource.matrixStride = program.GetActiveUniformMatrixStride(glIndex);
+                    resource.isRowMajor = program.GetActiveUniformIsRowMajor(glIndex);
+                    // A member of a named uniform block has no location, whatever the
+                    // frontend's own location table says (it hands one out to every uniform
+                    // so glUniform* can address block members through the global UBO).
+                    resource.location =
+                        resource.blockIndex >= 0 ? -1 : program.GetUniformLocation(refl.name);
+                }
+                model.uniforms.push_back(Move(resource));
+            }
+
+            // GL_ACTIVE_VARIABLES, both directions.
+            for (SizeT i = 0; i < model.uniforms.size(); ++i) {
+                const Resource& uniform = model.uniforms[i];
+                if (uniform.atomicCounterBufferIndex >= 0 &&
+                    uniform.atomicCounterBufferIndex < static_cast<GLint>(model.atomicCounterBuffers.size())) {
+                    model.atomicCounterBuffers[uniform.atomicCounterBufferIndex].activeVariables.push_back(
+                        static_cast<GLuint>(i));
+                }
+            }
+            for (SizeT blockIndex = 0; blockIndex < model.uniformBlocks.size(); ++blockIndex) {
+                // Members of an arrayed block are reflected once, against instance [0].
+                const Int owner = static_cast<Int>(program.GetUniformBlockMemberOwnerIndex(static_cast<Uint>(blockIndex)));
+                for (SizeT i = 0; i < model.uniforms.size(); ++i) {
+                    if (model.uniforms[i].blockIndex == owner) {
+                        model.uniformBlocks[blockIndex].activeVariables.push_back(static_cast<GLuint>(i));
+                    }
+                }
+            }
+            for (SizeT blockIndex = 0; blockIndex < model.storageBlocks.size(); ++blockIndex) {
+                for (SizeT i = 0; i < model.bufferVariables.size(); ++i) {
+                    if (model.bufferVariables[i].blockIndex == static_cast<GLint>(blockIndex)) {
+                        model.storageBlocks[blockIndex].activeVariables.push_back(static_cast<GLuint>(i));
+                    }
+                }
+            }
+        }
+
+        // A built-in interface block that a shader redeclares with fewer members keeps the
+        // omitted ones in its type when the redeclaration is ANONYMOUS - glslang hides them
+        // (basic type void) instead of erasing them, because the original shared declaration
+        // has to stay usable. Only the instance-named form erases. So a separable vertex
+        // program that redeclares `out gl_PerVertex { vec4 gl_Position; }` still carries
+        // gl_PointSize and gl_ClipDistance through the block-unwrapping reflection, and they
+        // are not part of its output interface.
+        Bool IsHiddenBlockMember(const glslang::TType* type) {
+            return type != nullptr && type->getBasicType() == glslang::EbtVoid;
+        }
+
+        void BuildStageIO(ProgramObject& program, const glslang::TProgram& reflection, Model& model) {
+            auto& mutableReflection = const_cast<glslang::TProgram&>(reflection);
+
+            const Int inputCount = mutableReflection.getNumPipeInputs();
+            for (Int index = 0; index < inputCount; ++index) {
+                const auto& refl = mutableReflection.getPipeInput(index);
+                const glslang::TType* type = refl.getType();
+                if (IsHiddenBlockMember(type)) continue;
+                Resource resource;
+                // The Vulkan-semantics parse reflects the vertex builtins under their SPIR-V
+                // names; GL enumerates the GL spellings.
+                const String& glName = ProgramObject::NormalizeBuiltinPipeInputName(refl.name);
+                resource.name = WithArraySuffix(glName, type);
+                resource.type = static_cast<GLenum>(refl.glDefineType);
+                resource.arraySize = ArraySizeOf(type, refl.size);
+                resource.location = program.GetAttributeLocation(refl.name);
+                if (resource.location < 0) resource.location = MappedLocation(static_cast<Int>(refl.layoutLocation()));
+                resource.isPerPatch = (type != nullptr && type->getQualifier().patch) ? 1 : 0;
+                resource.stages = static_cast<Uint32>(refl.stages);
+                model.programInputs.push_back(Move(resource));
+            }
+
+            // A color number, and therefore a color INDEX, exists only for a fragment stage's
+            // outputs. The output interface belongs to the program's last stage, so for a
+            // separable tessellation/geometry/vertex program these are varyings: asking the
+            // frag-data maps about them can still answer a location (a tess-control output
+            // carries its own layout(location=N)), and a location then manufactures a color
+            // index of 0 where GL requires -1
+            // (KHR-GL43.program_interface_query.separate-programs-tess-control).
+            const Bool lastStageIsFragment = mutableReflection.getIntermediate(EShLangFragment) != nullptr;
+            const Int outputCount = mutableReflection.getNumPipeOutputs();
+            for (Int index = 0; index < outputCount; ++index) {
+                const auto& refl = mutableReflection.getPipeOutput(index);
+                const glslang::TType* type = refl.getType();
+                if (IsHiddenBlockMember(type)) continue;
+                Resource resource;
+                resource.name = WithArraySuffix(refl.name, type);
+                resource.type = static_cast<GLenum>(refl.glDefineType);
+                resource.arraySize = ArraySizeOf(type, refl.size);
+                resource.location = MappedLocation(program.GetFragmentDataLocation(refl.name.c_str()));
+                if (resource.location < 0 || !lastStageIsFragment) {
+                    // A built-in output (gl_FragDepth, gl_SampleMask) has no location, and a
+                    // non-fragment stage's outputs have no color number at all - either way there
+                    // is no color index.
+                    resource.locationIndex = -1;
+                } else {
+                    resource.locationIndex = program.GetFragmentDataIndex(refl.name.c_str());
+                    // glBindFragDataLocationIndexed wins; otherwise the shader's
+                    // layout(index = N), which the frag-data maps never saw.
+                    if (resource.locationIndex == 0 && type != nullptr && type->getQualifier().hasIndex()) {
+                        resource.locationIndex = static_cast<GLint>(type->getQualifier().layoutIndex);
+                    }
+                }
+                resource.isPerPatch = (type != nullptr && type->getQualifier().patch) ? 1 : 0;
+                resource.stages = static_cast<Uint32>(refl.stages);
+                model.programOutputs.push_back(Move(resource));
+            }
+        }
+
+        void BuildXfb(ProgramObject& program, Model& model) {
+            const auto& requested = program.GetTransformFeedbackInterfaceNames();
+            const auto& captured = program.GetTransformFeedbackVaryings();
+            for (const String& name : requested) {
+                Resource resource;
+                resource.name = name;
+                // ARB_transform_feedback3's layout controls are enumerated as resources of
+                // type NONE: gl_NextBuffer with array size 0, gl_SkipComponentsN with N.
+                if (name == "gl_NextBuffer") {
+                    resource.type = GL_NONE;
+                    resource.arraySize = 0;
+                } else if (name.size() == 18 && name.compare(0, 17, "gl_SkipComponents") == 0 && name[17] >= '1' &&
+                           name[17] <= '4') {
+                    resource.type = GL_NONE;
+                    resource.arraySize = name[17] - '0';
+                } else {
+                    resource.type = GL_NONE;
+                    resource.arraySize = 1;
+                    for (const auto& varying : captured) {
+                        if (varying.name != name) continue;
+                        resource.type = varying.type;
+                        resource.arraySize = varying.size < 1 ? 1 : varying.size;
+                        resource.offset = static_cast<GLint>(varying.offsetBytes);
+                        resource.xfbBufferIndex = static_cast<GLint>(varying.bufferIndex);
+                        break;
+                    }
+                }
+                model.xfbVaryings.push_back(Move(resource));
+            }
+        }
+
+        Model BuildModel(ProgramObject& program) {
+            Model model;
+            if (!program.GetLinkStatus()) return model;
+            const glslang::TProgram* reflection = program.GetReflection();
+            if (reflection == nullptr) return model;
+            model.valid = true;
+
+            Vector<BlockKind> blockKind;
+            Vector<Int> blockInterfaceIndex;
+            BuildBlocks(program, *reflection, model, blockKind, blockInterfaceIndex);
+            BuildUniformsAndBufferVariables(program, *reflection, model, blockKind, blockInterfaceIndex);
+            BuildStageIO(program, *reflection, model);
+            BuildXfb(program, model);
+            return model;
+        }
+
+        const ResourceList& Select(const Model& model, GLenum programInterface) {
+            switch (programInterface) {
+            case GL_UNIFORM:
+                return model.uniforms;
+            case GL_UNIFORM_BLOCK:
+                return model.uniformBlocks;
+            case GL_ATOMIC_COUNTER_BUFFER:
+                return model.atomicCounterBuffers;
+            case GL_BUFFER_VARIABLE:
+                return model.bufferVariables;
+            case GL_SHADER_STORAGE_BLOCK:
+                return model.storageBlocks;
+            case GL_PROGRAM_INPUT:
+                return model.programInputs;
+            case GL_PROGRAM_OUTPUT:
+                return model.programOutputs;
+            case GL_TRANSFORM_FEEDBACK_VARYING:
+                return model.xfbVaryings;
+            default:
+                // The subroutine interfaces are accepted by the API but nothing can populate
+                // them: glslang refuses `subroutine` when generating SPIR-V, so a program
+                // using one never links. Zero active resources is the honest answer.
+                return EmptyList();
+            }
+        }
+    } // namespace
+
+    Bool IsInterfaceEnum(GLenum programInterface) {
+        switch (programInterface) {
+        case GL_UNIFORM:
+        case GL_UNIFORM_BLOCK:
+        case GL_PROGRAM_INPUT:
+        case GL_PROGRAM_OUTPUT:
+        case GL_BUFFER_VARIABLE:
+        case GL_SHADER_STORAGE_BLOCK:
+        case GL_ATOMIC_COUNTER_BUFFER:
+        case GL_TRANSFORM_FEEDBACK_VARYING:
+        case GL_TRANSFORM_FEEDBACK_BUFFER:
+        case GL_VERTEX_SUBROUTINE:
+        case GL_TESS_CONTROL_SUBROUTINE:
+        case GL_TESS_EVALUATION_SUBROUTINE:
+        case GL_GEOMETRY_SUBROUTINE:
+        case GL_FRAGMENT_SUBROUTINE:
+        case GL_COMPUTE_SUBROUTINE:
+        case GL_VERTEX_SUBROUTINE_UNIFORM:
+        case GL_TESS_CONTROL_SUBROUTINE_UNIFORM:
+        case GL_TESS_EVALUATION_SUBROUTINE_UNIFORM:
+        case GL_GEOMETRY_SUBROUTINE_UNIFORM:
+        case GL_FRAGMENT_SUBROUTINE_UNIFORM:
+        case GL_COMPUTE_SUBROUTINE_UNIFORM:
+            return true;
+        default:
+            return false;
+        }
+    }
+
+    Bool IsNamedInterface(GLenum programInterface) {
+        // GL 4.6 §7.3.1.2: the two buffer interfaces have no resource names, and asking for
+        // one is INVALID_ENUM (deliberately asymmetric with GetProgramInterfaceiv, which
+        // does count them).
+        return IsInterfaceEnum(programInterface) && programInterface != GL_ATOMIC_COUNTER_BUFFER &&
+               programInterface != GL_TRANSFORM_FEEDBACK_BUFFER;
+    }
+
+    Bool InterfaceHasLocations(GLenum programInterface) {
+        switch (programInterface) {
+        case GL_UNIFORM:
+        case GL_PROGRAM_INPUT:
+        case GL_PROGRAM_OUTPUT:
+        case GL_VERTEX_SUBROUTINE_UNIFORM:
+        case GL_TESS_CONTROL_SUBROUTINE_UNIFORM:
+        case GL_TESS_EVALUATION_SUBROUTINE_UNIFORM:
+        case GL_GEOMETRY_SUBROUTINE_UNIFORM:
+        case GL_FRAGMENT_SUBROUTINE_UNIFORM:
+        case GL_COMPUTE_SUBROUTINE_UNIFORM:
+            return true;
+        default:
+            return false;
+        }
+    }
+
+    Bool IsResourceProp(GLenum prop) {
+        switch (prop) {
+        case GL_NAME_LENGTH:
+        case GL_TYPE:
+        case GL_ARRAY_SIZE:
+        case GL_OFFSET:
+        case GL_BLOCK_INDEX:
+        case GL_ARRAY_STRIDE:
+        case GL_MATRIX_STRIDE:
+        case GL_IS_ROW_MAJOR:
+        case GL_ATOMIC_COUNTER_BUFFER_INDEX:
+        case GL_BUFFER_BINDING:
+        case GL_BUFFER_DATA_SIZE:
+        case GL_NUM_ACTIVE_VARIABLES:
+        case GL_ACTIVE_VARIABLES:
+        case GL_REFERENCED_BY_VERTEX_SHADER:
+        case GL_REFERENCED_BY_TESS_CONTROL_SHADER:
+        case GL_REFERENCED_BY_TESS_EVALUATION_SHADER:
+        case GL_REFERENCED_BY_GEOMETRY_SHADER:
+        case GL_REFERENCED_BY_FRAGMENT_SHADER:
+        case GL_REFERENCED_BY_COMPUTE_SHADER:
+        case GL_TOP_LEVEL_ARRAY_SIZE:
+        case GL_TOP_LEVEL_ARRAY_STRIDE:
+        case GL_LOCATION:
+        case GL_LOCATION_INDEX:
+        case GL_IS_PER_PATCH:
+        case GL_LOCATION_COMPONENT:
+        case GL_TRANSFORM_FEEDBACK_BUFFER_INDEX:
+        case GL_TRANSFORM_FEEDBACK_BUFFER_STRIDE:
+        case GL_NUM_COMPATIBLE_SUBROUTINES:
+        case GL_COMPATIBLE_SUBROUTINES:
+            return true;
+        default:
+            return false;
+        }
+    }
+
+    // GL 4.6 Table 7.2, transcribed row by row: which interfaces each property applies to.
+    // Too tight a table turns a currently-answered prop into a fresh INVALID_OPERATION, so
+    // the rows below are deliberately no narrower than the spec's.
+    Bool InterfaceSupportsProp(GLenum programInterface, GLenum prop) {
+        const Bool isSubroutine =
+            programInterface == GL_VERTEX_SUBROUTINE || programInterface == GL_TESS_CONTROL_SUBROUTINE ||
+            programInterface == GL_TESS_EVALUATION_SUBROUTINE || programInterface == GL_GEOMETRY_SUBROUTINE ||
+            programInterface == GL_FRAGMENT_SUBROUTINE || programInterface == GL_COMPUTE_SUBROUTINE;
+        const Bool isSubroutineUniform =
+            programInterface == GL_VERTEX_SUBROUTINE_UNIFORM ||
+            programInterface == GL_TESS_CONTROL_SUBROUTINE_UNIFORM ||
+            programInterface == GL_TESS_EVALUATION_SUBROUTINE_UNIFORM ||
+            programInterface == GL_GEOMETRY_SUBROUTINE_UNIFORM ||
+            programInterface == GL_FRAGMENT_SUBROUTINE_UNIFORM || programInterface == GL_COMPUTE_SUBROUTINE_UNIFORM;
+
+        switch (prop) {
+        case GL_NAME_LENGTH:
+            return programInterface != GL_ATOMIC_COUNTER_BUFFER && programInterface != GL_TRANSFORM_FEEDBACK_BUFFER;
+        case GL_TYPE:
+        case GL_ARRAY_SIZE:
+            return programInterface == GL_UNIFORM || programInterface == GL_PROGRAM_INPUT ||
+                   programInterface == GL_PROGRAM_OUTPUT || programInterface == GL_BUFFER_VARIABLE ||
+                   programInterface == GL_TRANSFORM_FEEDBACK_VARYING ||
+                   (prop == GL_ARRAY_SIZE && isSubroutineUniform);
+        case GL_OFFSET:
+            return programInterface == GL_UNIFORM || programInterface == GL_BUFFER_VARIABLE ||
+                   programInterface == GL_TRANSFORM_FEEDBACK_VARYING;
+        case GL_BLOCK_INDEX:
+        case GL_ARRAY_STRIDE:
+        case GL_MATRIX_STRIDE:
+        case GL_IS_ROW_MAJOR:
+            return programInterface == GL_UNIFORM || programInterface == GL_BUFFER_VARIABLE;
+        case GL_ATOMIC_COUNTER_BUFFER_INDEX:
+            return programInterface == GL_UNIFORM;
+        case GL_BUFFER_BINDING:
+        case GL_NUM_ACTIVE_VARIABLES:
+        case GL_ACTIVE_VARIABLES:
+            // Table 7.2 lists GL_TRANSFORM_FEEDBACK_BUFFER on these three rows too. This
+            // implementation enumerates no resources on that interface, so the query still
+            // ends in an error - but INVALID_VALUE for the out-of-range index, not the
+            // INVALID_OPERATION a narrower table would invent.
+            return programInterface == GL_UNIFORM_BLOCK || programInterface == GL_ATOMIC_COUNTER_BUFFER ||
+                   programInterface == GL_SHADER_STORAGE_BLOCK ||
+                   programInterface == GL_TRANSFORM_FEEDBACK_BUFFER;
+        case GL_BUFFER_DATA_SIZE:
+            return programInterface == GL_UNIFORM_BLOCK || programInterface == GL_ATOMIC_COUNTER_BUFFER ||
+                   programInterface == GL_SHADER_STORAGE_BLOCK;
+        case GL_REFERENCED_BY_VERTEX_SHADER:
+        case GL_REFERENCED_BY_TESS_CONTROL_SHADER:
+        case GL_REFERENCED_BY_TESS_EVALUATION_SHADER:
+        case GL_REFERENCED_BY_GEOMETRY_SHADER:
+        case GL_REFERENCED_BY_FRAGMENT_SHADER:
+        case GL_REFERENCED_BY_COMPUTE_SHADER:
+            return programInterface == GL_UNIFORM || programInterface == GL_UNIFORM_BLOCK ||
+                   programInterface == GL_ATOMIC_COUNTER_BUFFER || programInterface == GL_BUFFER_VARIABLE ||
+                   programInterface == GL_SHADER_STORAGE_BLOCK || programInterface == GL_PROGRAM_INPUT ||
+                   programInterface == GL_PROGRAM_OUTPUT || isSubroutineUniform;
+        case GL_TOP_LEVEL_ARRAY_SIZE:
+        case GL_TOP_LEVEL_ARRAY_STRIDE:
+            return programInterface == GL_BUFFER_VARIABLE;
+        case GL_LOCATION:
+            return InterfaceHasLocations(programInterface);
+        case GL_LOCATION_INDEX:
+            return programInterface == GL_PROGRAM_OUTPUT;
+        case GL_IS_PER_PATCH:
+        case GL_LOCATION_COMPONENT:
+            return programInterface == GL_PROGRAM_INPUT || programInterface == GL_PROGRAM_OUTPUT;
+        case GL_TRANSFORM_FEEDBACK_BUFFER_INDEX:
+            return programInterface == GL_TRANSFORM_FEEDBACK_VARYING;
+        case GL_TRANSFORM_FEEDBACK_BUFFER_STRIDE:
+            return programInterface == GL_TRANSFORM_FEEDBACK_BUFFER;
+        case GL_NUM_COMPATIBLE_SUBROUTINES:
+        case GL_COMPATIBLE_SUBROUTINES:
+            return isSubroutineUniform;
+        default:
+            (void)isSubroutine;
+            return false;
+        }
+    }
+
+    Int GetActiveResourceCount(ProgramObject& program, GLenum programInterface) {
+        const Model model = BuildModel(program);
+        return static_cast<Int>(Select(model, programInterface).size());
+    }
+
+    Int GetMaxNameLength(ProgramObject& program, GLenum programInterface) {
+        if (!IsNamedInterface(programInterface)) return 0;
+        const Model model = BuildModel(program);
+        SizeT longest = 0;
+        for (const Resource& resource : Select(model, programInterface)) {
+            longest = std::max(longest, resource.name.length() + 1);
+        }
+        return static_cast<Int>(longest);
+    }
+
+    Int GetMaxNumActiveVariables(ProgramObject& program, GLenum programInterface) {
+        const Model model = BuildModel(program);
+        SizeT longest = 0;
+        for (const Resource& resource : Select(model, programInterface)) {
+            longest = std::max(longest, resource.activeVariables.size());
+        }
+        return static_cast<Int>(longest);
+    }
+
+    GLuint GetResourceIndex(ProgramObject& program, GLenum programInterface, const char* name) {
+        if (name == nullptr || name[0] == '\0') return GL_INVALID_INDEX;
+        const Model model = BuildModel(program);
+        const ResourceList& resources = Select(model, programInterface);
+        const String query = name;
+        // The layout controls of an interleaved capture are enumerable but not addressable
+        // by name (GL 4.6 §7.3.1.1).
+        if (programInterface == GL_TRANSFORM_FEEDBACK_VARYING &&
+            (query == "gl_NextBuffer" ||
+             (query.size() == 18 && query.compare(0, 17, "gl_SkipComponents") == 0))) {
+            return GL_INVALID_INDEX;
+        }
+        for (SizeT i = 0; i < resources.size(); ++i) {
+            if (NamesMatch(resources[i].name, query)) return static_cast<GLuint>(i);
+        }
+        return GL_INVALID_INDEX;
+    }
+
+    Bool GetResourceName(ProgramObject& program, GLenum programInterface, GLuint index, String& outName) {
+        const Model model = BuildModel(program);
+        const ResourceList& resources = Select(model, programInterface);
+        if (index >= resources.size()) return false;
+        outName = resources[index].name;
+        return true;
+    }
+
+    Bool GetResourceProp(ProgramObject& program, GLenum programInterface, GLuint index, GLenum prop,
+                         Vector<GLint>& outValues) {
+        const Model model = BuildModel(program);
+        const ResourceList& resources = Select(model, programInterface);
+        if (index >= resources.size()) return false;
+        const Resource& resource = resources[index];
+
+        const auto referencedBy = [&resource](EShLanguage stage) {
+            return (resource.stages & static_cast<Uint32>(1u << stage)) != 0 ? GL_TRUE : GL_FALSE;
+        };
+
+        switch (prop) {
+        case GL_NAME_LENGTH:
+            outValues.push_back(static_cast<GLint>(resource.name.length() + 1));
+            break;
+        case GL_TYPE:
+            outValues.push_back(static_cast<GLint>(resource.type));
+            break;
+        case GL_ARRAY_SIZE:
+            outValues.push_back(resource.arraySize);
+            break;
+        case GL_OFFSET:
+            outValues.push_back(resource.offset);
+            break;
+        case GL_BLOCK_INDEX:
+            outValues.push_back(resource.blockIndex);
+            break;
+        case GL_ARRAY_STRIDE:
+            outValues.push_back(resource.arrayStride);
+            break;
+        case GL_MATRIX_STRIDE:
+            outValues.push_back(resource.matrixStride);
+            break;
+        case GL_IS_ROW_MAJOR:
+            outValues.push_back(resource.isRowMajor);
+            break;
+        case GL_ATOMIC_COUNTER_BUFFER_INDEX:
+            outValues.push_back(resource.atomicCounterBufferIndex);
+            break;
+        case GL_BUFFER_BINDING:
+            outValues.push_back(resource.bufferBinding);
+            break;
+        case GL_BUFFER_DATA_SIZE:
+            outValues.push_back(resource.bufferDataSize);
+            break;
+        case GL_NUM_ACTIVE_VARIABLES:
+            outValues.push_back(static_cast<GLint>(resource.activeVariables.size()));
+            break;
+        case GL_ACTIVE_VARIABLES:
+            for (const GLuint variable : resource.activeVariables) outValues.push_back(static_cast<GLint>(variable));
+            break;
+        case GL_REFERENCED_BY_VERTEX_SHADER:
+            outValues.push_back(referencedBy(EShLangVertex));
+            break;
+        case GL_REFERENCED_BY_TESS_CONTROL_SHADER:
+            outValues.push_back(referencedBy(EShLangTessControl));
+            break;
+        case GL_REFERENCED_BY_TESS_EVALUATION_SHADER:
+            outValues.push_back(referencedBy(EShLangTessEvaluation));
+            break;
+        case GL_REFERENCED_BY_GEOMETRY_SHADER:
+            outValues.push_back(referencedBy(EShLangGeometry));
+            break;
+        case GL_REFERENCED_BY_FRAGMENT_SHADER:
+            outValues.push_back(referencedBy(EShLangFragment));
+            break;
+        case GL_REFERENCED_BY_COMPUTE_SHADER:
+            outValues.push_back(referencedBy(EShLangCompute));
+            break;
+        case GL_TOP_LEVEL_ARRAY_SIZE:
+            outValues.push_back(resource.topLevelArraySize);
+            break;
+        case GL_TOP_LEVEL_ARRAY_STRIDE:
+            outValues.push_back(resource.topLevelArrayStride);
+            break;
+        case GL_LOCATION:
+            outValues.push_back(resource.location);
+            break;
+        case GL_LOCATION_INDEX:
+            outValues.push_back(resource.locationIndex);
+            break;
+        case GL_IS_PER_PATCH:
+            outValues.push_back(resource.isPerPatch);
+            break;
+        case GL_LOCATION_COMPONENT:
+            outValues.push_back(0);
+            break;
+        case GL_TRANSFORM_FEEDBACK_BUFFER_INDEX:
+            outValues.push_back(resource.xfbBufferIndex);
+            break;
+        default:
+            outValues.push_back(0);
+            break;
+        }
+        return true;
+    }
+
+    GLint GetResourceLocation(ProgramObject& program, GLenum programInterface, const char* name) {
+        if (name == nullptr || name[0] == '\0') return -1;
+        const String query = name;
+
+        String base;
+        Uint element = 0;
+        Bool malformed = false;
+        const Bool subscripted = SplitTrailingSubscript(query, base, element, malformed);
+        if (malformed) return -1;
+
+        const Model model = BuildModel(program);
+        const ResourceList& resources = Select(model, programInterface);
+        for (const Resource& resource : resources) {
+            if (NamesMatch(resource.name, query)) return resource.location;
+        }
+        if (!subscripted || element == 0) return -1;
+        // "d[1]" addresses the second element of an array resource enumerated as "d[0]".
+        for (const Resource& resource : resources) {
+            if (!NamesMatch(resource.name, base)) continue;
+            if (resource.location < 0 || static_cast<GLint>(element) >= resource.arraySize) return -1;
+            return resource.location + static_cast<GLint>(element);
+        }
+        return -1;
+    }
+
+    GLint GetResourceLocationIndex(ProgramObject& program, GLenum programInterface, const char* name) {
+        if (programInterface != GL_PROGRAM_OUTPUT || name == nullptr || name[0] == '\0') return -1;
+        const String query = name;
+        String base;
+        Uint element = 0;
+        Bool malformed = false;
+        const Bool subscripted = SplitTrailingSubscript(query, base, element, malformed);
+        if (malformed) return -1;
+
+        const Model model = BuildModel(program);
+        for (const Resource& resource : model.programOutputs) {
+            if (NamesMatch(resource.name, query)) return resource.locationIndex;
+        }
+        if (!subscripted) return -1;
+        for (const Resource& resource : model.programOutputs) {
+            if (!NamesMatch(resource.name, base)) continue;
+            if (resource.location < 0 || static_cast<GLint>(element) >= resource.arraySize) return -1;
+            return resource.locationIndex;
+        }
+        return -1;
+    }
+} // namespace MobileGL::MG_Impl::GLImpl::ProgramInterface

MobileGL/MG_Impl/GLImpl/Program/ProgramInterface.h

@@ -0,0 +1,66 @@
+// MobileGL - MobileGL/MG_Impl/GLImpl/Program/ProgramInterface.h
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#pragma once
+#include <Includes.h>
+
+namespace MobileGL::MG_State::GLState {
+    class ProgramObject;
+}
+
+// The GL program interface (ARB_program_interface_query / GL 4.3 §7.3.1) as a frontend
+// resource model.
+//
+// WHY IT IS HERE AND NOT IN A BACKEND. glGetProgramResource* describes the program the
+// APPLICATION wrote, in the application's namespace. Neither backend program is in that
+// namespace: DirectGLES compiles SPIRV-Cross-generated ESSL where default-block uniforms
+// live inside the synthesized MGL_GLOBAL_UBO (so a GL_UNIFORM location query against it is
+// structurally -1) and stage in/out names are rewritten; DirectVulkan has no GL-level
+// reflection at all and can only re-derive a partial, diverging copy. The one authoritative
+// source is the frontend glslang reflection a link already produced, which is the same
+// place glGetActiveUniform answers from. This layer generalizes that rule to every
+// interface, so the six entry points never consult gBackendFunctionsTable.
+//
+// NAMING RULES LIVE HERE, NOT IN ProgramObject. The interface query spells resources
+// differently from glGetActiveUniform / glGetActiveAttrib (an array is "name[0]", a lookup
+// accepts both "name" and "name[0]", a subscript must be a strict decimal). Those two
+// getters are what GL30-33 exercises and they must not move, so every normalization is
+// applied on the way in and out of THIS file.
+namespace MobileGL::MG_Impl::GLImpl::ProgramInterface {
+    using ProgramObject = MG_State::GLState::ProgramObject;
+
+    // <programInterface> is one of the GL 4.6 Table 7.1 interfaces.
+    Bool IsInterfaceEnum(GLenum programInterface);
+    // Interfaces whose resources have names (everything except GL_ATOMIC_COUNTER_BUFFER).
+    Bool IsNamedInterface(GLenum programInterface);
+    // <prop> is a property token GetProgramResourceiv knows at all (else GL_INVALID_ENUM).
+    Bool IsResourceProp(GLenum prop);
+    // <prop> applies to <programInterface> (else GL_INVALID_OPERATION).
+    Bool InterfaceSupportsProp(GLenum programInterface, GLenum prop);
+    // Interfaces GetProgramResourceLocation accepts (else GL_INVALID_ENUM).
+    Bool InterfaceHasLocations(GLenum programInterface);
+
+    // GL_ACTIVE_RESOURCES / GL_MAX_NAME_LENGTH / GL_MAX_NUM_ACTIVE_VARIABLES. All three
+    // report zero for an interface this implementation cannot enumerate and for a program
+    // that has not linked successfully - which is what the spec requires of a program with
+    // no active resources.
+    Int GetActiveResourceCount(ProgramObject& program, GLenum programInterface);
+    Int GetMaxNameLength(ProgramObject& program, GLenum programInterface);
+    Int GetMaxNumActiveVariables(ProgramObject& program, GLenum programInterface);
+
+    // GL_INVALID_INDEX when <name> names no active resource of the interface.
+    GLuint GetResourceIndex(ProgramObject& program, GLenum programInterface, const char* name);
+    // False when <index> is out of range for the interface (the caller raises INVALID_VALUE).
+    Bool GetResourceName(ProgramObject& program, GLenum programInterface, GLuint index, String& outName);
+    // Appends the value(s) of <prop> for the resource; GL_ACTIVE_VARIABLES appends several.
+    // False when <index> is out of range.
+    Bool GetResourceProp(ProgramObject& program, GLenum programInterface, GLuint index, GLenum prop,
+                         Vector<GLint>& outValues);
+    GLint GetResourceLocation(ProgramObject& program, GLenum programInterface, const char* name);
+    GLint GetResourceLocationIndex(ProgramObject& program, GLenum programInterface, const char* name);
+} // namespace MobileGL::MG_Impl::GLImpl::ProgramInterface

MobileGL/MG_Impl/GLImpl/Query/GL_Query.cpp

@@ -648,4 +648,39 @@ namespace MobileGL::MG_Impl::GLImpl {
         if (!ValidateQueryStreamIndex(__FUNCTION__, target, index)) return;
         GetQueryiv(target, pname, params);
     }
+
+    void DestroyAllQueryObjects() {
+        // Detach the registry under the lock, release outside it - same discipline
+        // (and the same accepted teardown race) as DestroyAllSyncObjects. Without
+        // this drain, every query the app left undeleted survived full library
+        // teardown in the process-global registry: the objects and their backend
+        // wrappers leaked across Destroy/Initialize cycles, stale ids kept
+        // answering IsQuery == GL_TRUE in the re-initialized library, and a later
+        // glDeleteQueries could hand the OLD backend's handle to a DIFFERENT
+        // backend's DeleteBackendQuery, which casts it to the wrong wrapper type.
+        UnorderedMap<GLuint, QueryObject*> orphans;
+        {
+            const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+            orphans.swap(g_liveQueryObjects);
+            g_activeTimeElapsedQueryId = 0;
+            g_activePrimitivesWrittenQueryId = 0;
+            g_activePrimitivesGeneratedQueryId = 0;
+            g_activeSamplesPassedQueryId = 0;
+        }
+        if (orphans.empty()) {
+            return;
+        }
+        // Backend handles must be released by the backend that created them, so
+        // this runs while the function table is still populated. Both backends'
+        // DeleteBackendQuery are generation-guarded, so a handle whose renderer
+        // or ES context is already gone frees only the wrapper.
+        const auto deleteBackendQuery = MG_Backend::gBackendFunctionsTable.GL.DeleteBackendQuery;
+        for (const auto& [_, queryObject] : orphans) {
+            if (deleteBackendQuery && queryObject->backendHandle) {
+                deleteBackendQuery(queryObject->backendHandle);
+            }
+            delete queryObject;
+        }
+        MGLOG_D("DestroyAllQueryObjects: reclaimed %zu query object(s) the app left undeleted", orphans.size());
+    }
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Query/GL_Query.h

@@ -29,4 +29,13 @@ namespace MobileGL::MG_Impl::GLImpl {
     void GetQueryBufferObjecti64v(GLuint id, GLuint buffer, GLenum pname, GLintptr offset);
     void GetQueryBufferObjectui64v(GLuint id, GLuint buffer, GLenum pname, GLintptr offset);
     void QueryCounter(GLuint id, GLenum target);
+    // Destroys every still-registered query object exactly as DeleteQueries would.
+    // GL requires queries to die with their context; called only from full library
+    // teardown (DestroyImpl), where no context survives on any thread, so the
+    // process-global registry can be drained wholesale. Must run while the backend
+    // function table is still populated: each backend handle has to be released by
+    // the backend that created it, never by a later re-initialized one (whose
+    // DeleteBackendQuery would cast the wrapper to the wrong backend's type).
+    // Same contract as DestroyAllSyncObjects.
+    void DestroyAllQueryObjects();
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/RenderState/GL_RenderState.cpp

@@ -8,6 +8,7 @@
 
 #include "GL_RenderState.h"
 #include <cmath>
+#include <MG_Impl/GLImpl/Getter/GL_Getter.h>
 #include <MG_State/GLState/Core.h>
 #include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
 #include <MG_Util/Converters/GLToMG/RenderStateEnumConverter.h>
@@ -19,28 +20,118 @@ namespace MobileGL::MG_Impl::GLImpl {
         return std::clamp(static_cast<Float>(value), 0.0f, 1.0f);
     }
 
-    static Bool ValidateIndexedBlendCapability(GLenum target, GLuint index, const char* functionName) {
+    // GL 4.6 core 17.3.2 and 22.1 give exactly two indexed capabilities: GL_BLEND, indexed by
-        if (target != GL_BLEND) {
+    // draw buffer, and GL_SCISSOR_TEST, indexed by viewport. They have DIFFERENT bounds
+    // (MAX_DRAW_BUFFERS vs MAX_VIEWPORTS), so the limit is picked per target rather than shared.
+    static Bool ValidateIndexedCapability(GLenum target, GLuint index, const char* functionName) {
+        GLuint limit = 0;
+        const char* indexName = nullptr;
+        switch (target) {
+        case GL_BLEND:
+            limit = MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS;
+            indexName = "Buffer";
+            break;
+        case GL_SCISSOR_TEST:
+            limit = RenderStateParameters::MAX_VIEWPORTS;
+            indexName = "Viewport";
+            break;
+        default:
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidEnum,
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
-                                             "Only GL_BLEND is supported for indexed capability state."));
+                                             "Only GL_BLEND and GL_SCISSOR_TEST are supported for indexed "
+                                             "capability state."));
             return false;
         }
 
-        if (index >= MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS) {
+        if (index >= limit) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue,
-                MakeUnique<GenericErrorInfo>(
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
-                    "MG_Impl/GLImpl", functionName,
+                                             String(indexName) + " index " + std::to_string(index) +
-                    "Buffer index " + std::to_string(index) + " is out of range. Max supported is " +
+                                                 " is out of range. Max supported is " + std::to_string(limit - 1) +
-                        std::to_string(MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS - 1) + "."));
+                                                 "."));
             return false;
         }
 
         return true;
     }
 
+    // ------------------ ARB_viewport_array parameter validation ------------------
+    // All three families share the same two shapes, so they share the two checkers. GL 4.6 core
+    // 13.6.1/17.3.2: an out-of-range index is GL_INVALID_VALUE, and so is a negative width or
+    // height. `first + count == MAX_VIEWPORTS` is LEGAL - only strictly greater is an error,
+    // which KHR-GL43.viewport_array.api_errors checks explicitly in both directions.
+    static Bool ValidateViewportIndex(GLuint index, const char* functionName) {
+        if (index < RenderStateParameters::MAX_VIEWPORTS) return true;
+
+        MG_State::pGLContext->RecordError(
+            ErrorCode::InvalidValue,
+            MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                         "Viewport index " + std::to_string(index) +
+                                             " is out of range. Max supported is " +
+                                             std::to_string(RenderStateParameters::MAX_VIEWPORTS - 1) + "."));
+        return false;
+    }
+
+    static Bool ValidateViewportRange(GLuint first, GLsizei count, const char* functionName) {
+        if (count < 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName, "count must not be negative."));
+            return false;
+        }
+        // Widened before adding: first is a GLuint and count a GLsizei, so `first + count` in
+        // 32 bits can wrap past MAX_VIEWPORTS and let an out-of-range range through.
+        const Uint64 last = static_cast<Uint64>(first) + static_cast<Uint64>(count);
+        if (last > RenderStateParameters::MAX_VIEWPORTS) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                             "first (" + std::to_string(first) + ") + count (" +
+                                                 std::to_string(count) + ") exceeds GL_MAX_VIEWPORTS (" +
+                                                 std::to_string(RenderStateParameters::MAX_VIEWPORTS) + ")."));
+            return false;
+        }
+        return true;
+    }
+
+    template <typename T>
+    static Bool ValidateNonNegativeExtent(T width, T height, const char* functionName) {
+        if (width >= T(0) && height >= T(0)) return true;
+
+        MG_State::pGLContext->RecordError(
+            ErrorCode::InvalidValue,
+            MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName, "Width and height must be non-negative."));
+        return false;
+    }
+
+    // The array forms are all-or-nothing: one bad element rejects the whole call with a SINGLE
+    // GL_INVALID_VALUE and leaves every rectangle untouched. api_errors relies on both halves -
+    // it passes a full 16-element array with exactly one negative extent and then asserts the
+    // error queue holds exactly one entry.
+    template <typename T>
+    static Bool ValidateArrayExtents(GLsizei count, const T* v, const char* functionName) {
+        for (GLsizei i = 0; i < count; ++i) {
+            if (v[i * 4 + 2] >= T(0) && v[i * 4 + 3] >= T(0)) continue;
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                             "Width and height must be non-negative (element " + std::to_string(i) +
+                                                 ")."));
+            return false;
+        }
+        return true;
+    }
+
+    static Bool ValidateNonNullArray(const void* v, const char* functionName) {
+        if (v != nullptr) return true;
+        MG_State::pGLContext->RecordError(
+            ErrorCode::InvalidValue,
+            MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName, "value pointer cannot be null."));
+        return false;
+    }
+
     static Bool TryConvertBlendEquation(GLenum mode, const char* functionName,
                                         ::MobileGL::BlendEquation& outEquation) {
         outEquation = MG_Util::ConvertGLEnumToBlendEquation(mode);
@@ -92,16 +183,70 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void Viewport_State(GLint x, GLint y, GLsizei width, GLsizei height) {
-        if (width < 0 || height < 0) {
+        if (!ValidateNonNegativeExtent(width, height, "Viewport_State")) return;
-            MG_State::pGLContext->RecordError(ErrorCode::InvalidValue,
-                                              MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "Viewport_State",
-                                                                           "Width abd height must be non-negative."));
-            return;
-        }
 
         MG_State::pGLContext->SetViewport(IntVec4(x, y, width, height));
     }
 
+    // ------------------ ARB_viewport_array setters ------------------
+    void ViewportArrayv_State(GLuint first, GLsizei count, const GLfloat* v) {
+        if (!ValidateViewportRange(first, count, "ViewportArrayv_State")) return;
+        if (count == 0) return;
+        if (!ValidateNonNullArray(v, "ViewportArrayv_State")) return;
+        if (!ValidateArrayExtents(count, v, "ViewportArrayv_State")) return;
+
+        for (GLsizei i = 0; i < count; ++i) {
+            MG_State::pGLContext->SetViewportIndexed(first + static_cast<GLuint>(i),
+                                                     FloatVec4(v[i * 4 + 0], v[i * 4 + 1], v[i * 4 + 2], v[i * 4 + 3]));
+        }
+    }
+
+    void ViewportIndexedf_State(GLuint index, GLfloat x, GLfloat y, GLfloat w, GLfloat h) {
+        if (!ValidateViewportIndex(index, "ViewportIndexedf_State")) return;
+        if (!ValidateNonNegativeExtent(w, h, "ViewportIndexedf_State")) return;
+
+        MG_State::pGLContext->SetViewportIndexed(index, FloatVec4(x, y, w, h));
+    }
+
+    void ScissorArrayv_State(GLuint first, GLsizei count, const GLint* v) {
+        if (!ValidateViewportRange(first, count, "ScissorArrayv_State")) return;
+        if (count == 0) return;
+        if (!ValidateNonNullArray(v, "ScissorArrayv_State")) return;
+        if (!ValidateArrayExtents(count, v, "ScissorArrayv_State")) return;
+
+        for (GLsizei i = 0; i < count; ++i) {
+            MG_State::pGLContext->SetScissorBoxIndexed(first + static_cast<GLuint>(i),
+                                                       IntVec4(v[i * 4 + 0], v[i * 4 + 1], v[i * 4 + 2], v[i * 4 + 3]));
+        }
+    }
+
+    void ScissorIndexed_State(GLuint index, GLint left, GLint bottom, GLsizei width, GLsizei height) {
+        if (!ValidateViewportIndex(index, "ScissorIndexed_State")) return;
+        if (!ValidateNonNegativeExtent(width, height, "ScissorIndexed_State")) return;
+
+        MG_State::pGLContext->SetScissorBoxIndexed(index, IntVec4(left, bottom, width, height));
+    }
+
+    void DepthRangeArrayv_State(GLuint first, GLsizei count, const GLdouble* v) {
+        if (!ValidateViewportRange(first, count, "DepthRangeArrayv_State")) return;
+        if (count == 0) return;
+        if (!ValidateNonNullArray(v, "DepthRangeArrayv_State")) return;
+
+        for (GLsizei i = 0; i < count; ++i) {
+            MG_State::pGLContext->SetDepthRangeIndexed(
+                first + static_cast<GLuint>(i),
+                FloatVec2(ClampUnitFloat(static_cast<GLfloat>(v[i * 2 + 0])),
+                          ClampUnitFloat(static_cast<GLfloat>(v[i * 2 + 1]))));
+        }
+    }
+
+    void DepthRangeIndexed_State(GLuint index, GLdouble n, GLdouble f) {
+        if (!ValidateViewportIndex(index, "DepthRangeIndexed_State")) return;
+
+        MG_State::pGLContext->SetDepthRangeIndexed(
+            index, FloatVec2(ClampUnitFloat(static_cast<GLfloat>(n)), ClampUnitFloat(static_cast<GLfloat>(f))));
+    }
+
     void StencilOpSeparate_State(GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass) {
         Bool applyFront = false;
         Bool applyBack = false;
@@ -174,12 +319,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void Scissor_State(GLint x, GLint y, GLsizei width, GLsizei height) {
-        if (width < 0 || height < 0) {
+        if (!ValidateNonNegativeExtent(width, height, "Scissor_State")) return;
-            MG_State::pGLContext->RecordError(ErrorCode::InvalidValue,
-                                              MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "Scissor_State",
-                                                                           "Width abd height must be non-negative."));
-            return;
-        }
 
         MG_State::pGLContext->SetScissorBox(IntVec4(x, y, width, height));
     }
@@ -335,7 +475,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     GLboolean IsEnabledi_State(GLenum target, GLuint index) {
-        if (!ValidateIndexedBlendCapability(target, index, "IsEnabledi_State")) {
+        if (!ValidateIndexedCapability(target, index, "IsEnabledi_State")) {
             return GL_FALSE;
         }
 
@@ -380,7 +520,25 @@ namespace MobileGL::MG_Impl::GLImpl {
             return;
         }
 
+        // GL 4.6 core 22.1: glGetBooleani_v answers EVERY indexed state, not just the indexed
+        // capabilities - a non-boolean value simply reads back as "is it non-zero". Routing the
+        // non-capability enums to the pname table glGetIntegeri_v already owns is what makes
+        // that true; without it a query like glGetBooleani_v(GL_MAX_COMPUTE_WORK_GROUP_COUNT, 0)
+        // came back GL_INVALID_ENUM (KHR-GL43.compute_shader.max).
+        if (MG_Util::ConvertGLEnumToCapabilityInput(target) != CapabilityInput::Unknown) {
             *data = IsEnabledi_State(target, index);
+            return;
+        }
+        GLint values[4] = {};
+        GetIntegeri_v(target, index, values);
+        // The ARB_viewport_array rectangles are the only multi-component indexed state that
+        // reaches here; writing element 0 alone would leave the caller's other three untouched.
+        const GLsizei components = target == GL_VIEWPORT || target == GL_SCISSOR_BOX
+            ? 4
+            : (target == GL_DEPTH_RANGE ? 2 : 1);
+        for (GLsizei i = 0; i < components; ++i) {
+            data[i] = values[i] != 0 ? GL_TRUE : GL_FALSE;
+        }
     }
 
     GLboolean IsEnabled_State(GLenum cap) {
@@ -713,7 +871,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void Disablei_State(GLenum target, GLuint index) {
-        if (!ValidateIndexedBlendCapability(target, index, "Disablei_State")) {
+        if (!ValidateIndexedCapability(target, index, "Disablei_State")) {
             return;
         }
 
@@ -731,7 +889,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void Enablei_State(GLenum target, GLuint index) {
-        if (!ValidateIndexedBlendCapability(target, index, "Enablei_State")) {
+        if (!ValidateIndexedCapability(target, index, "Enablei_State")) {
             return;
         }
 
@@ -785,6 +943,44 @@ namespace MobileGL::MG_Impl::GLImpl {
         Viewport_State(x, y, width, height);
     }
 
+    void ViewportArrayv(GLuint first, GLsizei count, const GLfloat* v) {
+        ViewportArrayv_State(first, count, v);
+    }
+
+    void ViewportIndexedf(GLuint index, GLfloat x, GLfloat y, GLfloat w, GLfloat h) {
+        ViewportIndexedf_State(index, x, y, w, h);
+    }
+
+    void ViewportIndexedfv(GLuint index, const GLfloat* v) {
+        // The index is validated before the pointer is touched: glViewportIndexedfv(MAX, nullptr)
+        // must be one GL_INVALID_VALUE, not a null dereference.
+        if (!ValidateViewportIndex(index, "ViewportIndexedfv")) return;
+        if (!ValidateNonNullArray(v, "ViewportIndexedfv")) return;
+        ViewportIndexedf_State(index, v[0], v[1], v[2], v[3]);
+    }
+
+    void ScissorArrayv(GLuint first, GLsizei count, const GLint* v) {
+        ScissorArrayv_State(first, count, v);
+    }
+
+    void ScissorIndexed(GLuint index, GLint left, GLint bottom, GLsizei width, GLsizei height) {
+        ScissorIndexed_State(index, left, bottom, width, height);
+    }
+
+    void ScissorIndexedv(GLuint index, const GLint* v) {
+        if (!ValidateViewportIndex(index, "ScissorIndexedv")) return;
+        if (!ValidateNonNullArray(v, "ScissorIndexedv")) return;
+        ScissorIndexed_State(index, v[0], v[1], v[2], v[3]);
+    }
+
+    void DepthRangeArrayv(GLuint first, GLsizei count, const GLdouble* v) {
+        DepthRangeArrayv_State(first, count, v);
+    }
+
+    void DepthRangeIndexed(GLuint index, GLdouble n, GLdouble f) {
+        DepthRangeIndexed_State(index, n, f);
+    }
+
     void StencilOpSeparate(GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass) {
         StencilOpSeparate_State(face, sfail, dpfail, dppass);
     }

MobileGL/MG_Impl/GLImpl/RenderState/GL_RenderState.h

@@ -20,6 +20,16 @@ namespace MobileGL::MG_Impl::GLImpl {
     void Enablei(GLenum target, GLuint index);
     void BlendFunc(GLenum sfactor, GLenum dfactor);
     void Viewport(GLint x, GLint y, GLsizei width, GLsizei height);
+    // ARB_viewport_array (core since GL 4.1). Every one of these addresses the same 16-element
+    // indexed state the classic glViewport/glScissor/glDepthRange trio broadcasts to.
+    void ViewportArrayv(GLuint first, GLsizei count, const GLfloat* v);
+    void ViewportIndexedf(GLuint index, GLfloat x, GLfloat y, GLfloat w, GLfloat h);
+    void ViewportIndexedfv(GLuint index, const GLfloat* v);
+    void ScissorArrayv(GLuint first, GLsizei count, const GLint* v);
+    void ScissorIndexed(GLuint index, GLint left, GLint bottom, GLsizei width, GLsizei height);
+    void ScissorIndexedv(GLuint index, const GLint* v);
+    void DepthRangeArrayv(GLuint first, GLsizei count, const GLdouble* v);
+    void DepthRangeIndexed(GLuint index, GLdouble n, GLdouble f);
     void StencilOpSeparate(GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass);
     void StencilOp(GLenum fail, GLenum zfail, GLenum zpass);
     void StencilMaskSeparate(GLenum face, GLuint mask);

MobileGL/MG_Impl/GLImpl/Sampler/GL_Sampler.cpp

@@ -8,6 +8,8 @@
 
 #include "GL_Sampler.h"
 #include "Validators.h"
+#include "../Getter/GL_Getter.h"
+#include "../Texture/GL_Texture.h"
 #include <MG_State/GLState/Core.h>
 #include <MG_Util/Converters/GLToMG/TextureEnumConverter.h>
 #include <MG_Util/Converters/MGToGL/TextureEnumConverter.h>
@@ -268,9 +270,18 @@ namespace MobileGL::MG_Impl::GLImpl {
         }
     }
 
+    // The number of texture units a sampler may be bound to is the same count a TEXTURE may be
+    // bound to - GL 3.3 core 3.8.2 names GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS for both - so it is
+    // computed once, in GetCombinedTextureImageUnitCount, and named here for the sampler-side
+    // readers below. Two copies of that arithmetic is how glBindSamplers and glBindTextures would
+    // come to disagree about which units exist.
+    static GLint GetSamplerBindableTextureUnitCount() {
+        return GetCombinedTextureImageUnitCount();
+    }
+
     void BindSampler_State(GLuint unit, GLuint sampler) {
         MGLOG_D("BindSampler_State: unit = %u, sampler = %u", unit, sampler);
-        if (unit >= MG_State::GLState::TextureState::MAX_TEXTURE_IMAGE_UNITS) {
+        if (static_cast<Uint64>(unit) >= static_cast<Uint64>(GetSamplerBindableTextureUnitCount())) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue,
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "BindSampler", "texture unit out of range"));
@@ -309,9 +320,37 @@ namespace MobileGL::MG_Impl::GLImpl {
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "BindSamplers", "count must be non-negative"));
             return;
         }
+        // ARB_multi_bind: the whole [first, first + count) range is checked up front and a
+        // range that runs past the last texture unit is INVALID_OPERATION - not the
+        // INVALID_VALUE the single-bind BindSampler_State reports per element, and nothing is
+        // bound when it fails. Both gates read the same limit (see
+        // GetSamplerBindableTextureUnitCount), so an out-of-range multi-bind can no longer slip
+        // past this check and be caught one element at a time with the wrong error class.
+        const GLint maxTextureUnits = GetSamplerBindableTextureUnitCount();
+        if (static_cast<Uint64>(first) + static_cast<Uint64>(count) > static_cast<Uint64>(maxTextureUnits)) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "BindSamplers",
+                                             "first + count exceeds the number of texture units."));
+            return;
+        }
 
+        // ARB_multi_bind adds one rule the single-bind path does not have: "samplers will not be
+        // created if they do not exist", so a name that is not an existing sampler OBJECT is
+        // INVALID_OPERATION here (KHR-GL44.multi_bind.errors_bind_samplers). Per element, not
+        // all-or-nothing - the extension defines glBindSamplers as a loop, so a bad entry costs
+        // its own texture unit and leaves the rest of the range bound.
         for (GLsizei i = 0; i < count; ++i) {
-            BindSampler_State(first + i, samplers ? samplers[i] : 0);
+            const GLuint sampler = samplers ? samplers[i] : 0;
+            if (sampler != 0 && !MG_State::pGLContext->ValidateSamplerObject(sampler)) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidOperation,
+                    MakeUnique<GenericErrorInfo>(
+                        "MG_Impl/GLImpl", "BindSamplers",
+                        std::format("samplers[{}] ({}) is not the name of an existing sampler object.", i, sampler)));
+                continue;
+            }
+            BindSampler_State(first + i, sampler);
         }
     }
 

MobileGL/MG_Impl/GLImpl/Sampler/Validators.cpp

@@ -75,7 +75,11 @@ namespace MobileGL::MG_Impl::GLImpl::SamplerImpl {
             break;
 
         case GL_TEXTURE_COMPARE_FUNC:
-            if (param < GL_LEQUAL || param > GL_ALWAYS) {
+            // The eight depth-compare functions are contiguous from GL_NEVER (0x0200) to
+            // GL_ALWAYS (0x0207); GL_LEQUAL sits in the middle of that block, so starting
+            // the range there rejected NEVER/LESS/EQUAL and let GREATER/NOTEQUAL/GEQUAL
+            // through only by accident of them being above LEQUAL.
+            if (param < GL_NEVER || param > GL_ALWAYS) {
                 MG_State::pGLContext->RecordError(ErrorCode::InvalidEnum,
                                                   MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateSamplerParam",
                                                                                "Invalid compare function parameter"));

MobileGL/MG_Impl/GLImpl/Texture/GL_Texture.h

@@ -37,6 +37,8 @@ namespace MobileGL::MG_Impl::GLImpl {
                            GLenum format, GLenum type, const void* pixels);
     void TextureSubImage3D(GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width,
                            GLsizei height, GLsizei depth, GLenum format, GLenum type, const void* pixels);
+    void CompressedTextureSubImage2D(GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLsizei width,
+                                     GLsizei height, GLenum format, GLsizei imageSize, const void* data);
     void TextureParameterf(GLuint texture, GLenum pname, GLfloat param);
     void TextureParameterfv(GLuint texture, GLenum pname, const GLfloat* params);
     void TextureParameteri(GLuint texture, GLenum pname, GLint param);
@@ -132,5 +134,11 @@ namespace MobileGL::MG_Impl::GLImpl {
     void CompressedTexImage1D(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLint border,
                               GLsizei imageSize, const void* data);
     void BindTexture(GLenum target, GLuint texture);
+    void BindTextures(GLuint first, GLsizei count, const GLuint* textures);
+    void BindImageTextures(GLuint first, GLsizei count, const GLuint* textures);
     void ActiveTexture(GLenum texture);
+    // The number of texture image units a texture or a sampler may be bound to: what the backend
+    // advertises as GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS, clamped by the frontend's fixed unit-array
+    // capacity. Shared so the texture and sampler multi-bind range checks cannot drift apart.
+    GLint GetCombinedTextureImageUnitCount();
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Texture/Validators.cpp

@@ -15,6 +15,7 @@
 #include <MG_Util/Converters/MGToGL/TextureEnumConverter.h>
 #include <MG_Util/Converters/MGToMG/TextureEnumConverter.h>
 #include <MG_Util/Converters/MGToStr/TextureEnumConverter.h>
+#include <MG_Util/Metrics/TextureMetrics.h>
 
 namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
     Bool ValidateTextureTarget(TextureTarget target) {
@@ -353,7 +354,64 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
         return true;
     }
 
-    Bool ValidateTextureObject(SharedPtr<MG_State::GLState::ITextureObject> textureObject) {
+    Bool ValidateTextureLevelExists(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject, Int level,
+                                    const char* caller) {
+        // A null object is somebody else's error to report - ValidateTextureObject runs
+        // first at every call site and has already recorded it.
+        if (!textureObject) return false;
+
+        const auto* mipmapTexture = MG_State::GLState::AsMipmapTexture(textureObject.get());
+        if (mipmapTexture == nullptr) {
+            // The only non-mipmap storage class is a buffer texture, and GL_TEXTURE_BUFFER is
+            // not a target glCopyImageSubData accepts at all (it is in the CTS's invalid-target
+            // set). Declining here is not the error code the spec asks for - that would be
+            // INVALID_ENUM from a target check this validator is not - but it does keep a
+            // texture with no image levels whatsoever from reaching a backend that would
+            // dereference a backend texture it never created.
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", caller,
+                                             "Texture has no mipmap levels to address."));
+            return false;
+        }
+
+        // What this number is, exactly, because two other things are almost it and neither is
+        // safe to assume: it is the number of level SLOTS the shadow has allocated - holes
+        // included, since MipmapStorage::AllocateLevel grows to level+1 and never fills the gap.
+        // For a cube map MipmapUploadTargetArray reports face +X's chain rather than the union.
+        //
+        // The guarantee that matters is one-sided: this count is always >= the level count the
+        // backends derive (VkTextureManager::GetUploadMipLevelCount stops at the first level
+        // with a non-positive extent, so it can only be shorter). That is the safe direction -
+        // no copy to a level the texture genuinely has is ever rejected here. It is NOT an
+        // exact match, so the backends keep their own range guard for the band in between: a
+        // chain with a hole (level 0 and 2 defined, 1 not) is accepted by this predicate and
+        // declined by the backend, which is a silent no-op rather than a copy. That band is a
+        // backend storage limitation, not a validation one - rejecting it here with
+        // INVALID_VALUE would be refusing a copy the spec permits.
+        const Uint levelCount = mipmapTexture->GetMipmapLevelCount();
+
+        if (levelCount == 0) {
+            // No image has ever been defined on this texture, so the fault is the texture,
+            // not the number: GL 4.6 core 18.3.2 asks for INVALID_OPERATION when an object a
+            // copy names is an incomplete texture.
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", caller,
+                                             "Texture has no image defined at any level."));
+            return false;
+        }
+        if (level < 0 || static_cast<Uint>(level) >= levelCount) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", caller,
+                                             "Texture level does not exist in this texture."));
+            return false;
+        }
+        return true;
+    }
+
+    Bool ValidateTextureObject(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject) {
         if (!textureObject) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
@@ -376,7 +434,7 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
         return true;
     }
 
-    Bool ValidateTextureTargetUniformity(SharedPtr<MG_State::GLState::ITextureObject> textureObject,
+    Bool ValidateTextureTargetUniformity(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject,
                                          TextureTarget target) {
         if (!textureObject) return true; // should be created later
         TextureTarget prevTarget = textureObject->GetTarget();
@@ -390,7 +448,7 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
         return true;
     }
 
-    Bool ValidateTextureSubImageOffsets(SharedPtr<MG_State::GLState::ITextureObject> textureObject, Int xoffset,
+    Bool ValidateTextureSubImageOffsets(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject, Int xoffset,
                                         Int width, Int yoffset, Int height, Int zoffset, Int depth) {
         auto baseSize = textureObject->GetBaseSize();
         if (xoffset < 0 || (xoffset + width) > baseSize.x()) {
@@ -424,19 +482,141 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
         return true;
     }
 
-    Bool ValidateBaseInternalFormatMatch(TextureInternalFormat format1, TextureInternalFormat format2) {
+    namespace {
-        auto unsizedFormat1 = MG_Util::ConvertInternalFormatToUnsized(format1);
+        // Component set of an UNSIZED base internal format, as the bitmask GL 4.6 SS 8.6
-        auto unsizedFormat2 = MG_Util::ConvertInternalFormatToUnsized(format2);
+        // reasons about. Colour components are independent bits so "subset" is a plain
-        if (unsizedFormat1 != unsizedFormat2) {
+        // mask test; depth and stencil are their own components and never satisfy a
+        // colour request (or each other).
+        enum : Uint32 {
+            kComponentR = 1u << 0,
+            kComponentG = 1u << 1,
+            kComponentB = 1u << 2,
+            kComponentA = 1u << 3,
+            kComponentDepth = 1u << 4,
+            kComponentStencil = 1u << 5,
+        };
+
+        Uint32 BaseFormatComponents(TextureInternalFormat unsizedFormat) {
+            switch (unsizedFormat) {
+            case TextureInternalFormat::Red:
+                return kComponentR;
+            case TextureInternalFormat::RG:
+                return kComponentR | kComponentG;
+            case TextureInternalFormat::RGB:
+                return kComponentR | kComponentG | kComponentB;
+            case TextureInternalFormat::RGBA:
+                return kComponentR | kComponentG | kComponentB | kComponentA;
+            case TextureInternalFormat::DepthComponent:
+                return kComponentDepth;
+            case TextureInternalFormat::DepthStencil:
+                return kComponentDepth | kComponentStencil;
+            default:
+                return 0;
+            }
+        }
+    } // namespace
+
+    CopyImageTexelBlock ResolveCopyImageTexelBlock(TextureInternalFormat format, GLenum compressedFormat) {
+        CopyImageTexelBlock block{};
+        if (compressedFormat != GL_NONE) {
+            const auto info = MG_Util::GetCompressedFormatInfo(compressedFormat);
+            if (info.blockByteSize != 0) {
+                block.byteSize = info.blockByteSize;
+                block.blockWidth = info.blockWidth;
+                block.blockHeight = info.blockHeight;
+                block.compressed = true;
+                return block;
+            }
+        }
+        // The size MobileGL actually stores a texel of this format in, which for every format GL
+        // gives a required size is that required size. The handful of legacy formats GL leaves
+        // implementation-defined (R3_G3_B2, RGB4/5/10/12, RGBA2/12) have no view class in table
+        // 8.22 to be compared against anyway, and this is the size that decides whether a raw
+        // copy between them would in fact preserve the bytes.
+        block.byteSize = MG_Util::GetSizedInternalFormatSizeInBytes(format);
+        return block;
+    }
+
+    Bool ValidateCopyImageFormatCompatibility(const CopyImageTexelBlock& srcBlock,
+                                              const CopyImageTexelBlock& dstBlock) {
+        if (srcBlock.byteSize == 0 || dstBlock.byteSize == 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateCopyImageFormatCompatibility",
+                                             "A copied image has no storage whose texel size is known."));
+            return false;
+        }
+        if (srcBlock.byteSize != dstBlock.byteSize) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
                 MakeUnique<GenericErrorInfo>(
-                    std::format("MG_Impl/GLImpl", "ValidateBaseInternalFormatMatch",
+                    "MG_Impl/GLImpl", "ValidateCopyImageFormatCompatibility",
-                                "The base internal format of the two formats do not match ({} vs. {})",
+                    std::format("The two images' texel blocks are different sizes ({} vs. {} bytes), so the "
-                                MG_Util::ConvertTextureInternalFormatToString(unsizedFormat1).c_str(),
+                                "formats are not copy-compatible.",
-                                MG_Util::ConvertTextureInternalFormatToString(unsizedFormat2).c_str())));
+                                srcBlock.byteSize, dstBlock.byteSize)));
+            return false;
+        }
+        // Two compressed images additionally have to agree on the SHAPE of the block, not only
+        // its size: an 8-byte 4x4 block and a hypothetical 8-byte 8x8 one hold different texel
+        // counts, and GL 4.6 core 18.3.2 requires both dimensions to match.
+        if (srcBlock.compressed && dstBlock.compressed &&
+            (srcBlock.blockWidth != dstBlock.blockWidth || srcBlock.blockHeight != dstBlock.blockHeight)) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl", "ValidateCopyImageFormatCompatibility",
+                    std::format("The two compressed images have different block dimensions ({}x{} vs. {}x{}).",
+                                srcBlock.blockWidth, srcBlock.blockHeight, dstBlock.blockWidth,
+                                dstBlock.blockHeight)));
             return false;
         }
         return true;
-    } // namespace TextureImpl
+    }
+
+    Bool ValidateCopyImageBlockAlignment(const CopyImageTexelBlock& block, Int x, Int y, Int width, Int height,
+                                         Int imageWidth, Int imageHeight, const char* endpointName) {
+        if (!block.compressed) return true;
+        const Int blockWidth = static_cast<Int>(block.blockWidth);
+        const Int blockHeight = static_cast<Int>(block.blockHeight);
+        if (blockWidth <= 1 && blockHeight <= 1) return true;
+        // The origin is unconditional; the extent gets the "or it reaches the edge of the image"
+        // exemption GL 4.6 core 18.3.2 grants, which is what lets a 16x16 BPTC image be copied
+        // whole even when the last block is partial.
+        const Bool originAligned = (x % blockWidth == 0) && (y % blockHeight == 0);
+        const Bool widthOk = (width % blockWidth == 0) || (x + width == imageWidth);
+        const Bool heightOk = (height % blockHeight == 0) || (y + height == imageHeight);
+        if (originAligned && widthOk && heightOk) return true;
+        MG_State::pGLContext->RecordError(
+            ErrorCode::InvalidValue,
+            MakeUnique<GenericErrorInfo>(
+                "MG_Impl/GLImpl", "ValidateCopyImageBlockAlignment",
+                std::format("The {} region [{}, {}] + [{} x {}] is not aligned to the {}x{} compressed block "
+                            "grid of a {} x {} image.",
+                            endpointName, x, y, width, height, blockWidth, blockHeight, imageWidth, imageHeight)));
+        return false;
+    }
+
+    Bool ValidateCopyTexImageBaseFormatSubset(TextureInternalFormat destFormat, TextureInternalFormat srcFormat) {
+        const auto unsizedDest = MG_Util::ConvertInternalFormatToUnsized(destFormat);
+        const auto unsizedSrc = MG_Util::ConvertInternalFormatToUnsized(srcFormat);
+        // GL 4.6 SS 8.6: glCopyTexImage* may request a SUBSET of the read buffer's components,
+        // not an exact match - GL_RGB from an RGBA8 framebuffer is textbook legal and is what
+        // Minecraft and its mods do. glCopyTexImage2D used to run the exact-match predicate
+        // above and turn its rejection into an uncaught exception through the C GL ABI, so the
+        // app died rather than seeing a GL error.
+        const Uint32 destComponents = BaseFormatComponents(unsizedDest);
+        const Uint32 srcComponents = BaseFormatComponents(unsizedSrc);
+        if (destComponents == 0 || srcComponents == 0 || (destComponents & ~srcComponents) != 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl", "ValidateCopyTexImageBaseFormatSubset",
+                    std::format("the read buffer's base internal format {} does not provide every component of "
+                                "the requested internal format {}",
+                                MG_Util::ConvertTextureInternalFormatToString(unsizedSrc),
+                                MG_Util::ConvertTextureInternalFormatToString(unsizedDest))));
+            return false;
+        }
+        return true;
+    }
 } // namespace MobileGL::MG_Impl::GLImpl::TextureImpl

MobileGL/MG_Impl/GLImpl/Texture/Validators.h

@@ -30,15 +30,53 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
                                                           TextureInternalFormat internalFormat,
                                                           TexturePixelDataType type);
     Bool ValidateTextureLevelWithUploadTarget(TextureUploadTarget target, Int level);
-    Bool ValidateTextureObject(SharedPtr<MG_State::GLState::ITextureObject> textureObject);
+    // "Is <level> a level this texture actually has?", which ValidateTextureLevelNumber above
+    // does NOT answer - that one only bounds the index by GL_MAX_TEXTURE_SIZE and knows nothing
+    // about the object. Entry points that resolve a level straight into a backend image
+    // subresource need this one: a level the texture never had is GL_INVALID_VALUE (GL 4.6 core
+    // 18.3.2), and passing it through instead reaches the driver as an out-of-range subresource.
+    // Note the error split is per-entry-point, so this is not universally reusable:
+    // glClearTexImage owes INVALID_OPERATION for the same out-of-range level and spells its own
+    // copy of this predicate in GL_Texture.cpp (GetClearTextureObject).
+    Bool ValidateTextureLevelExists(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject, Int level,
+                                    const char* caller);
+    Bool ValidateTextureObject(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject);
     // Rejects the per-target default texture objects (name 0) with GL_INVALID_OPERATION for entry
     // points that require a GenTextures-created texture, e.g. TexStorage* ("An INVALID_OPERATION
     // error is generated if zero is bound to target", ARB_texture_storage).
     Bool ValidateTextureNotDefault(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject,
                                    const char* caller);
-    Bool ValidateTextureTargetUniformity(SharedPtr<MG_State::GLState::ITextureObject> textureObject,
+    Bool ValidateTextureTargetUniformity(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject,
                                          TextureTarget target);
-    Bool ValidateTextureSubImageOffsets(SharedPtr<MG_State::GLState::ITextureObject> textureObject, Int xoffset,
+    Bool ValidateTextureSubImageOffsets(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject, Int xoffset,
                                         Int width, Int yoffset = 0, Int height = 0, Int zoffset = 0, Int depth = 0);
-    Bool ValidateBaseInternalFormatMatch(TextureInternalFormat format1, TextureInternalFormat format2);
+    // The texel block of one glCopyImageSubData endpoint, resolved to the two things the
+    // compatibility rule actually asks about. `compressed` is not redundant with a block bigger
+    // than 1x1: it is what distinguishes "compressed, and so the region is measured in texels of
+    // a blocked image" from "uncompressed, and so it is measured in texels".
+    struct CopyImageTexelBlock {
+        SizeT byteSize = 0;
+        Uint blockWidth = 1;
+        Uint blockHeight = 1;
+        Bool compressed = false;
+    };
+    // `compressedFormat` is the GLenum a glCompressedTexImage* upload recorded for the level, or
+    // GL_NONE. It has to be asked for separately because MobileGL stores every compressed format
+    // in uncompressed storage (ConvertGLEnumToTextureInternalFormat), so the TextureInternalFormat
+    // alone can no longer tell a BPTC image from the RGBA8 backing it.
+    CopyImageTexelBlock ResolveCopyImageTexelBlock(TextureInternalFormat format, GLenum compressedFormat);
+    // GL 4.6 core 18.3.2: the two images must be COMPATIBLE, and compatible means their texel
+    // blocks are the same SIZE - not that they share a base internal format. RGBA32UI into
+    // RGBA32F is legal (both 128-bit) while RGBA8 into RGBA32F is not, and a compressed image
+    // pairs with an uncompressed one whose texel is as big as the compressed block.
+    Bool ValidateCopyImageFormatCompatibility(const CopyImageTexelBlock& srcBlock,
+                                              const CopyImageTexelBlock& dstBlock);
+    // GL 4.6 core 18.3.2: for a compressed image the region's origin must sit on a block
+    // boundary and its size must be a whole number of blocks - unless the edge it runs to is
+    // the edge of the image.
+    Bool ValidateCopyImageBlockAlignment(const CopyImageTexelBlock& block, Int x, Int y, Int width, Int height,
+                                         Int imageWidth, Int imageHeight, const char* endpointName);
+    // GL 4.6 SS 8.6 subset rule for glCopyTexImage*: the read buffer must supply every component
+    // the requested internalformat asks for, but may supply more.
+    Bool ValidateCopyTexImageBaseFormatSubset(TextureInternalFormat destFormat, TextureInternalFormat srcFormat);
 } // namespace MobileGL::MG_Impl::GLImpl::TextureImpl

MobileGL/MG_Impl/GLImpl/VertexArray/GL_VertexArray.cpp

@@ -8,6 +8,7 @@
 
 #include "GL_VertexArray.h"
 #include "Validators.h"
+#include <MG_Backend/BackendObjects.h>
 #include <MG_Impl/GLImpl/Buffer/Validators.h>
 #include <MG_State/GLState/Core.h>
 #include <MG_State/GLState/ErrorState/Error.h>
@@ -105,6 +106,21 @@ namespace MobileGL::MG_Impl::GLImpl {
             return pname == GL_CURRENT_VERTEX_ATTRIB;
         }
 
+        // The two ARB_vertex_attrib_binding per-attribute queries. They do not live on the
+        // resolved VertexAttribute (which is the flat, already-combined view) but on the VAO's
+        // binding-point mapping, so they need the object, not the attribute.
+        static bool TryGetVertexAttribBindingQuery(GLuint index, GLenum pname, GLint& out) {
+            if (pname != GL_VERTEX_ATTRIB_BINDING && pname != GL_VERTEX_ATTRIB_RELATIVE_OFFSET) return false;
+            const auto& vao = MG_State::pGLContext->GetBoundVertexArray();
+            if (!vao) {
+                out = 0;
+                return true;
+            }
+            out = pname == GL_VERTEX_ATTRIB_BINDING ? static_cast<GLint>(vao->GetAttributeBindingIndex(index))
+                                                    : static_cast<GLint>(vao->GetAttributeRelativeOffset(index));
+            return true;
+        }
+
         // The stride a pointer-style call gives its binding point: the argument when it is non-zero,
         // otherwise the tightly packed element size (GL 4.6 core 10.3.2). A packed 2_10_10_10 or
         // 10F_11F_11F attribute is one 32-bit word regardless of its component count.
@@ -163,6 +179,28 @@ namespace MobileGL::MG_Impl::GLImpl {
             return vao;
         }
 
+        // The ARB_vertex_attrib_binding entry points that take no vertex array name modify the
+        // *bound* vertex array, and in a core profile the default vertex array (name 0) is not
+        // one: every one of them is INVALID_OPERATION there (GL 4.6 core 10.3.1, and the tail of
+        // each KHR-GL4x.vertex_attrib_binding.negative-* case checks exactly this). MobileGL
+        // keeps a real object at name 0 for the compatibility paths, so GetBoundVertexArray
+        // never returns null and the rule has to be spelled out - behind the same gate the VAO-0
+        // draw rule already uses (MOBILEGL_RELAXED_SEMANTICS, plus "the context never asked for
+        // a core profile"), so applications that legitimately run relaxed keep working.
+        static SharedPtr<MG_State::GLState::VertexArrayObject> GetBoundVertexArrayForBindingApi(const char* funcName) {
+            auto vao = GetBoundVertexArrayOrError(funcName);
+            if (!vao) return nullptr;
+            if (vao->GetExternalIndex() == 0 && !MG_State::IsRelaxedSemanticsActive()) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidOperation,
+                    MakeUnique<GenericErrorInfo>(
+                        "MG_Impl/GLImpl", funcName,
+                        "The default vertex array object cannot be modified in a core profile."));
+                return nullptr;
+            }
+            return vao;
+        }
+
         static bool ValidateVertexAttribPname(GLenum pname) {
             switch (pname) {
             case GL_VERTEX_ATTRIB_ARRAY_ENABLED:
@@ -173,8 +211,16 @@ namespace MobileGL::MG_Impl::GLImpl {
             case GL_CURRENT_VERTEX_ATTRIB:
             case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING:
             case GL_VERTEX_ATTRIB_ARRAY_INTEGER:
+            // Core since GL 4.1 (ARB_vertex_attrib_64bit). It was rejected while no attribute could
+            // ever be long; now that IsLong is real state the pname has to be accepted.
+            case GL_VERTEX_ATTRIB_ARRAY_LONG:
             case GL_VERTEX_ATTRIB_ARRAY_DIVISOR:
             case GL_VERTEX_ATTRIB_ARRAY_POINTER:
+            // ARB_vertex_attrib_binding (core since GL 4.3). The binding-point view is real
+            // state on the VAO (GetAttributeBindingIndex / GetAttributeRelativeOffset), so
+            // both of its per-attribute queries are answerable.
+            case GL_VERTEX_ATTRIB_BINDING:
+            case GL_VERTEX_ATTRIB_RELATIVE_OFFSET:
                 return true;
             default:
                 MG_State::pGLContext->RecordError(
@@ -269,9 +315,10 @@ namespace MobileGL::MG_Impl::GLImpl {
 
         auto offset = reinterpret_cast<SizeT>(pointer);
 
-        vao->SetAttributeFormat(index, size, dataType, false, stride, offset, true, false);
+        const int effectiveStride = EffectiveVertexStride(stride, size, type);
+        vao->SetAttributeFormat(index, size, dataType, false, stride, offset, true, false, effectiveStride);
         vao->BindAttributeBuffer(index, vbo);
-        vao->MirrorPointerIntoBinding(index, vbo, offset, EffectiveVertexStride(stride, size, type));
+        vao->MirrorPointerIntoBinding(index, vbo, offset, effectiveStride);
     }
 
     void VertexAttribPointer_State(GLuint index, GLint size, GLenum type, GLboolean normalized, GLsizei stride,
@@ -299,9 +346,11 @@ namespace MobileGL::MG_Impl::GLImpl {
         // backend can pick the reversed VkFormat / pass GL_BGRA through to a GLES driver.
         const bool isBgra = (size == static_cast<GLint>(GL_BGRA));
         const int effectiveSize = isBgra ? 4 : size;
-        vao->SetAttributeFormat(index, effectiveSize, dataType, normalized, stride, offset, false, isBgra);
+        const int effectiveStride = EffectiveVertexStride(stride, effectiveSize, type);
+        vao->SetAttributeFormat(index, effectiveSize, dataType, normalized, stride, offset, false, isBgra,
+                                effectiveStride);
         vao->BindAttributeBuffer(index, vbo);
-        vao->MirrorPointerIntoBinding(index, vbo, offset, EffectiveVertexStride(stride, effectiveSize, type));
+        vao->MirrorPointerIntoBinding(index, vbo, offset, effectiveStride);
     }
 
     void BindVertexArray_State(GLuint array) {
@@ -460,19 +509,38 @@ namespace MobileGL::MG_Impl::GLImpl {
                                         relativeoffset, isBgra);
     }
 
-    // The long (64-bit) attribute format. MobileGL has no 64-bit vertex attributes, so nothing is
+    // The long (64-bit) attribute format: the values reach the shader as doubles, unconverted
-    // recorded; what the entry point owes the application is the parameter validation, which is
+    // (GL 4.6 core 10.3.2). ValidateVertexAttribLFormat has already pinned type to GL_DOUBLE, so the
-    // observable through glGetError regardless of whether the format could be used in a draw.
+    // recorded DataType is always Float64 - what IsLong adds is that this is the *unconverted* form,
-    static void VertexAttribLFormatSeparate_State(GLuint attribindex, GLint size, GLenum type,
+    // as opposed to VertexAttribFormat(GL_DOUBLE), which asks for a float conversion.
+    //
+    // Whether the backend can feed it is detected, not assumed: DirectVulkan needs shaderFloat64,
+    // and DirectGLES can never have it at all. A backend without it declines here, loudly - GL error
+    // plus a log line naming the reason - rather than accepting state no draw could honour and
+    // rendering garbage. The matching startup POST row is in MG_Util/SelfTest/DriverPost.cpp.
+    static void VertexAttribLFormatSeparate_State(const SharedPtr<MG_State::GLState::VertexArrayObject>& vao,
+                                                  GLuint attribindex, GLint size, GLenum type,
                                                   GLuint relativeoffset) {
         if (!VertexArrayImpl::ValidateVertexAttributeIndex(attribindex)) return;
         if (!VertexArrayImpl::ValidateVertexAttribLFormat(attribindex, size, type)) return;
         if (!VertexArrayImpl::ValidateVertexAttribRelativeOffset(relativeoffset)) return;
 
+        if (!MG_Backend::pActiveBackendObject ||
+            !MG_Backend::pActiveBackendObject->GetDynamicParameters().SupportsFloat64VertexAttributes) {
+            MGLOG_W_ONCE("VertexAttribLFormat: attribute %u asked for a 64-bit (GL_DOUBLE) format, but this "
+                    "backend has no double-precision vertex attribute support - see the "
+                    "\"64-bit vertex attributes\" / \"shaderFloat64\" POST row for what that costs",
+                    attribindex);
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "VertexAttribLFormat",
-                                         "64-bit vertex attributes are not supported."));
+                                             "64-bit vertex attributes are not supported by this backend."));
+            return;
+        }
+
+        vao->SetAttributeFormatSeparate(attribindex, size, MG_Util::ConvertGLEnumToDataType(type),
+                                        /*normalized: */ false, /*isInteger: */ false, relativeoffset,
+                                        /*isBgra: */ false, /*isLong: */ true);
     }
 
     void VertexArrayAttribFormat_State(GLuint vaobj, GLuint attribindex, GLint size, GLenum type,
@@ -901,7 +969,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             params[0] = static_cast<GLfloat>(attr->Size);
             return;
         case GL_VERTEX_ATTRIB_ARRAY_STRIDE:
-            params[0] = static_cast<GLfloat>(attr->Stride);
+            params[0] = static_cast<GLfloat>(attr->LegacyStride);
             return;
         case GL_VERTEX_ATTRIB_ARRAY_TYPE:
             params[0] = static_cast<GLfloat>(MG_Util::ConvertDataTypeToGLEnum(attr->Type));
@@ -915,9 +983,19 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_VERTEX_ATTRIB_ARRAY_INTEGER:
             params[0] = attr->IsInteger ? 1.0f : 0.0f;
             return;
+        case GL_VERTEX_ATTRIB_ARRAY_LONG:
+            params[0] = attr->IsLong ? 1.0f : 0.0f;
+            return;
         case GL_VERTEX_ATTRIB_ARRAY_DIVISOR:
             params[0] = static_cast<GLfloat>(attr->Divisor);
             return;
+        case GL_VERTEX_ATTRIB_BINDING:
+        case GL_VERTEX_ATTRIB_RELATIVE_OFFSET: {
+            GLint value = 0;
+            TryGetVertexAttribBindingQuery(index, pname, value);
+            params[0] = static_cast<GLfloat>(value);
+            return;
+        }
         default:
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidEnum,
@@ -961,7 +1039,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             params[0] = static_cast<GLdouble>(attr->Size);
             return;
         case GL_VERTEX_ATTRIB_ARRAY_STRIDE:
-            params[0] = static_cast<GLdouble>(attr->Stride);
+            params[0] = static_cast<GLdouble>(attr->LegacyStride);
             return;
         case GL_VERTEX_ATTRIB_ARRAY_TYPE:
             params[0] = static_cast<GLdouble>(MG_Util::ConvertDataTypeToGLEnum(attr->Type));
@@ -975,9 +1053,19 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_VERTEX_ATTRIB_ARRAY_INTEGER:
             params[0] = attr->IsInteger ? 1.0 : 0.0;
             return;
+        case GL_VERTEX_ATTRIB_ARRAY_LONG:
+            params[0] = attr->IsLong ? 1.0 : 0.0;
+            return;
         case GL_VERTEX_ATTRIB_ARRAY_DIVISOR:
             params[0] = static_cast<GLdouble>(attr->Divisor);
             return;
+        case GL_VERTEX_ATTRIB_BINDING:
+        case GL_VERTEX_ATTRIB_RELATIVE_OFFSET: {
+            GLint value = 0;
+            TryGetVertexAttribBindingQuery(index, pname, value);
+            params[0] = static_cast<GLdouble>(value);
+            return;
+        }
         default:
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidEnum,
@@ -1016,8 +1104,11 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_VERTEX_ATTRIB_ARRAY_SIZE:
             params[0] = attr->Size;
             return;
+        // The legacy shadow, not the resolved draw stride: GL 4.6 core table 23.3 defines this
+        // as the last glVertexAttrib*Pointer argument, which glBindVertexBuffer must not
+        // overwrite even though it does overwrite what the backend actually reads.
         case GL_VERTEX_ATTRIB_ARRAY_STRIDE:
-            params[0] = attr->Stride;
+            params[0] = attr->LegacyStride;
             return;
         case GL_VERTEX_ATTRIB_ARRAY_TYPE:
             params[0] = static_cast<GLint>(MG_Util::ConvertDataTypeToGLEnum(attr->Type));
@@ -1031,9 +1122,16 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_VERTEX_ATTRIB_ARRAY_INTEGER:
             params[0] = attr->IsInteger ? GL_TRUE : GL_FALSE;
             return;
+        case GL_VERTEX_ATTRIB_ARRAY_LONG:
+            params[0] = attr->IsLong ? GL_TRUE : GL_FALSE;
+            return;
         case GL_VERTEX_ATTRIB_ARRAY_DIVISOR:
             params[0] = static_cast<GLint>(attr->Divisor);
             return;
+        case GL_VERTEX_ATTRIB_BINDING:
+        case GL_VERTEX_ATTRIB_RELATIVE_OFFSET:
+            TryGetVertexAttribBindingQuery(index, pname, params[0]);
+            return;
         default:
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidEnum,
@@ -1068,7 +1166,7 @@ namespace MobileGL::MG_Impl::GLImpl {
         }
 
         const auto& attr = vao->GetAttribute(index);
-        *pointer = reinterpret_cast<void*>(attr.Offset);
+        *pointer = reinterpret_cast<void*>(attr.LegacyPointer);
     }
 
     void GetVertexAttribIiv(GLuint index, GLenum pname, GLint* params) {
@@ -1152,7 +1250,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             *param = static_cast<GLint>(attr.Size);
             return;
         case GL_VERTEX_ATTRIB_ARRAY_STRIDE:
-            *param = static_cast<GLint>(attr.Stride);
+            *param = static_cast<GLint>(attr.LegacyStride);
             return;
         case GL_VERTEX_ATTRIB_ARRAY_TYPE:
             *param = static_cast<GLint>(MG_Util::ConvertDataTypeToGLEnum(attr.Type));
@@ -1164,8 +1262,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             *param = attr.IsInteger ? GL_TRUE : GL_FALSE;
             return;
         case GL_VERTEX_ATTRIB_ARRAY_LONG:
-            // 64-bit attributes are not supported, so no attribute is ever a long one.
+            *param = attr.IsLong ? GL_TRUE : GL_FALSE;
-            *param = GL_FALSE;
             return;
         case GL_VERTEX_ATTRIB_ARRAY_DIVISOR:
             *param = static_cast<GLint>(attr.Divisor);
@@ -1173,6 +1270,9 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_VERTEX_ATTRIB_RELATIVE_OFFSET:
             *param = static_cast<GLint>(vao->GetAttributeRelativeOffset(index));
             return;
+        case GL_VERTEX_ATTRIB_BINDING:
+            *param = static_cast<GLint>(vao->GetAttributeBindingIndex(index));
+            return;
         default:
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidEnum,
@@ -1222,14 +1322,14 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void BindVertexBuffer(GLuint bindingindex, GLuint buffer, GLintptr offset, GLsizei stride) {
-        auto vao = GetBoundVertexArrayOrError("BindVertexBuffer");
+        auto vao = GetBoundVertexArrayForBindingApi("BindVertexBuffer");
         if (!vao) return;
         VertexBufferBinding_State(vao, bindingindex, buffer, offset, stride, "BindVertexBuffer");
     }
 
     void BindVertexBuffers(GLuint first, GLsizei count, const GLuint* buffers, const GLintptr* offsets,
                            const GLsizei* strides) {
-        auto vao = GetBoundVertexArrayOrError("BindVertexBuffers");
+        auto vao = GetBoundVertexArrayForBindingApi("BindVertexBuffers");
         if (!vao) return;
         if (!ValidateVertexBindingRange(first, count, "BindVertexBuffers")) return;
         for (GLsizei i = 0; i < count; ++i) {
@@ -1243,33 +1343,33 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void VertexAttribFormat(GLuint attribindex, GLint size, GLenum type, GLboolean normalized, GLuint relativeoffset) {
-        auto vao = GetBoundVertexArrayOrError("VertexAttribFormat");
+        auto vao = GetBoundVertexArrayForBindingApi("VertexAttribFormat");
         if (!vao) return;
         VertexAttribFormatSeparate_State(vao, attribindex, size, type, normalized, relativeoffset, false,
                                          "VertexAttribFormat");
     }
 
     void VertexAttribIFormat(GLuint attribindex, GLint size, GLenum type, GLuint relativeoffset) {
-        auto vao = GetBoundVertexArrayOrError("VertexAttribIFormat");
+        auto vao = GetBoundVertexArrayForBindingApi("VertexAttribIFormat");
         if (!vao) return;
         VertexAttribFormatSeparate_State(vao, attribindex, size, type, GL_FALSE, relativeoffset, true,
                                          "VertexAttribIFormat");
     }
 
     void VertexAttribLFormat(GLuint attribindex, GLint size, GLenum type, GLuint relativeoffset) {
-        auto vao = GetBoundVertexArrayOrError("VertexAttribLFormat");
+        auto vao = GetBoundVertexArrayForBindingApi("VertexAttribLFormat");
         if (!vao) return;
-        VertexAttribLFormatSeparate_State(attribindex, size, type, relativeoffset);
+        VertexAttribLFormatSeparate_State(vao, attribindex, size, type, relativeoffset);
     }
 
     void VertexArrayAttribLFormat(GLuint vaobj, GLuint attribindex, GLint size, GLenum type, GLuint relativeoffset) {
         auto vao = GetNamedVertexArrayObject_State(vaobj, "VertexArrayAttribLFormat");
         if (!vao) return;
-        VertexAttribLFormatSeparate_State(attribindex, size, type, relativeoffset);
+        VertexAttribLFormatSeparate_State(vao, attribindex, size, type, relativeoffset);
     }
 
     void VertexAttribBinding(GLuint attribindex, GLuint bindingindex) {
-        auto vao = GetBoundVertexArrayOrError("VertexAttribBinding");
+        auto vao = GetBoundVertexArrayForBindingApi("VertexAttribBinding");
         if (!vao) return;
         if (!VertexArrayImpl::ValidateVertexAttributeIndex(attribindex)) return;
         if (!ValidateVertexBindingIndex(bindingindex, "VertexAttribBinding")) return;
@@ -1277,7 +1377,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void VertexBindingDivisor(GLuint bindingindex, GLuint divisor) {
-        auto vao = GetBoundVertexArrayOrError("VertexBindingDivisor");
+        auto vao = GetBoundVertexArrayForBindingApi("VertexBindingDivisor");
         if (!vao) return;
         if (!ValidateVertexBindingIndex(bindingindex, "VertexBindingDivisor")) return;
         vao->SetBindingDivisor(bindingindex, divisor);

MobileGL/MG_Impl/GLImpl/VertexArray/Validators.cpp

@@ -147,6 +147,30 @@ namespace MobileGL::MG_Impl::GLImpl::VertexArrayImpl {
             return false;
         }
 
+        // The integer path takes exactly the six signed/unsigned integer types (GL 4.6
+        // core 10.3.2): BYTE, UNSIGNED_BYTE, SHORT, UNSIGNED_SHORT, INT, UNSIGNED_INT.
+        // A blacklist could not express that: GL_FLOAT, GL_HALF_FLOAT,
+        // GL_DOUBLE and GL_FIXED all convert to a perfectly valid DataType, so they slipped
+        // through and were recorded as integer attributes.
+        if (integerPath) {
+            switch (type) {
+            case DataType::Int8:
+            case DataType::Uint8:
+            case DataType::Int16:
+            case DataType::Uint16:
+            case DataType::Int32:
+            case DataType::Uint32:
+                break;
+            default:
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidEnum,
+                    MakeUnique<GenericErrorInfo>(
+                        "MG_Impl/GLImpl", fn,
+                        std::format("Type is not an integer vertex attribute type (attribute {}).", index)));
+                return false;
+            }
+        }
+
         if (sizeRaw == static_cast<GLint>(GL_BGRA)) {
             // GL_BGRA is a float-path-only size: it needs GL_UNSIGNED_BYTE or a 2_10_10_10 type and
             // normalized == GL_TRUE. On the integer path it is simply an out-of-range size.

MobileGL/MG_Impl/GLXImpl/Exporting/Definitions.cpp

@@ -16,3 +16,256 @@ MOBILEGL_GLX_API void* glXGetProcAddress(const char* name) {
 MOBILEGL_GLX_API void* glXGetProcAddressARB(const char* name) {
     return MG_Impl::GLXImpl::GetProcAddressARB(name);
 }
+
+#if defined(__linux__) && !defined(__ANDROID__)
+#include "../GLXImpl.h"
+
+namespace GLXImpl = MobileGL::MG_Impl::GLXImpl;
+
+// GLX handle/type spellings from GL/glx.h, expressed without including it:
+// GLXContext/GLXFBConfig are opaque pointers, drawables are XIDs, Bool is int,
+// and XVisualInfo* crosses as void*.
+
+MOBILEGL_GLX_API int glXQueryExtension(Display* dpy, int* errorBase, int* eventBase) {
+    return GLXImpl::QueryExtension(dpy, errorBase, eventBase);
+}
+
+MOBILEGL_GLX_API int glXQueryVersion(Display* dpy, int* major, int* minor) {
+    return GLXImpl::QueryVersion(dpy, major, minor);
+}
+
+MOBILEGL_GLX_API const char* glXQueryExtensionsString(Display* dpy, int screen) {
+    return GLXImpl::QueryExtensionsString(dpy, screen);
+}
+
+MOBILEGL_GLX_API const char* glXGetClientString(Display* dpy, int name) {
+    return GLXImpl::GetClientString(dpy, name);
+}
+
+MOBILEGL_GLX_API const char* glXQueryServerString(Display* dpy, int screen, int name) {
+    return GLXImpl::QueryServerString(dpy, screen, name);
+}
+
+MOBILEGL_GLX_API void** glXGetFBConfigs(Display* dpy, int screen, int* nelements) {
+    return GLXImpl::GetFBConfigs(dpy, screen, nelements);
+}
+
+MOBILEGL_GLX_API void** glXChooseFBConfig(Display* dpy, int screen, const int* attribList,
+                                          int* nelements) {
+    return GLXImpl::ChooseFBConfig(dpy, screen, attribList, nelements);
+}
+
+MOBILEGL_GLX_API int glXGetFBConfigAttrib(Display* dpy, void* config, int attribute, int* value) {
+    return GLXImpl::GetFBConfigAttrib(dpy, config, attribute, value);
+}
+
+MOBILEGL_GLX_API void* glXGetVisualFromFBConfig(Display* dpy, void* config) {
+    return GLXImpl::GetVisualFromFBConfig(dpy, config);
+}
+
+MOBILEGL_GLX_API void* glXChooseVisual(Display* dpy, int screen, int* attribList) {
+    return GLXImpl::ChooseVisual(dpy, screen, attribList);
+}
+
+MOBILEGL_GLX_API int glXGetConfig(Display* dpy, void* visualInfo, int attribute, int* value) {
+    return GLXImpl::GetConfig(dpy, visualInfo, attribute, value);
+}
+
+MOBILEGL_GLX_API void* glXCreateContext(Display* dpy, void* visualInfo, void* shareList, int direct) {
+    return GLXImpl::CreateContext(dpy, visualInfo, shareList, direct);
+}
+
+MOBILEGL_GLX_API void* glXCreateNewContext(Display* dpy, void* config, int renderType,
+                                           void* shareList, int direct) {
+    return GLXImpl::CreateNewContext(dpy, config, renderType, shareList, direct);
+}
+
+MOBILEGL_GLX_API void* glXCreateContextAttribsARB(Display* dpy, void* config, void* shareContext,
+                                                  int direct, const int* attribList) {
+    return GLXImpl::CreateContextAttribsARB(dpy, config, shareContext, direct, attribList);
+}
+
+MOBILEGL_GLX_API void glXDestroyContext(Display* dpy, void* context) {
+    GLXImpl::DestroyContext(dpy, context);
+}
+
+MOBILEGL_GLX_API int glXMakeCurrent(Display* dpy, unsigned long drawable, void* context) {
+    return GLXImpl::MakeCurrent(dpy, drawable, context);
+}
+
+MOBILEGL_GLX_API int glXMakeContextCurrent(Display* dpy, unsigned long draw, unsigned long read,
+                                           void* context) {
+    return GLXImpl::MakeContextCurrent(dpy, draw, read, context);
+}
+
+MOBILEGL_GLX_API void glXSwapBuffers(Display* dpy, unsigned long drawable) {
+    GLXImpl::SwapBuffers(dpy, drawable);
+}
+
+MOBILEGL_GLX_API unsigned long glXCreateWindow(Display* dpy, void* config, unsigned long window,
+                                               const int* attribList) {
+    return GLXImpl::CreateWindow(dpy, config, window, attribList);
+}
+
+MOBILEGL_GLX_API void glXDestroyWindow(Display* dpy, unsigned long window) {
+    GLXImpl::DestroyWindow(dpy, window);
+}
+
+MOBILEGL_GLX_API void* glXGetCurrentContext() {
+    return GLXImpl::GetCurrentContext();
+}
+
+MOBILEGL_GLX_API unsigned long glXGetCurrentDrawable() {
+    return GLXImpl::GetCurrentDrawable();
+}
+
+MOBILEGL_GLX_API unsigned long glXGetCurrentReadDrawable() {
+    return GLXImpl::GetCurrentReadDrawable();
+}
+
+MOBILEGL_GLX_API Display* glXGetCurrentDisplay() {
+    return GLXImpl::GetCurrentDisplay();
+}
+
+MOBILEGL_GLX_API int glXIsDirect(Display* dpy, void* context) {
+    return GLXImpl::IsDirect(dpy, context);
+}
+
+MOBILEGL_GLX_API void glXWaitGL() {
+    GLXImpl::WaitGL();
+}
+
+MOBILEGL_GLX_API void glXWaitX() {
+    GLXImpl::WaitX();
+}
+
+MOBILEGL_GLX_API int glXQueryContext(Display* dpy, void* context, int attribute, int* value) {
+    return GLXImpl::QueryContext(dpy, context, attribute, value);
+}
+
+MOBILEGL_GLX_API void glXQueryDrawable(Display* dpy, unsigned long drawable, int attribute,
+                                       unsigned int* value) {
+    GLXImpl::QueryDrawable(dpy, drawable, attribute, value);
+}
+
+MOBILEGL_GLX_API void glXSwapIntervalEXT(Display* dpy, unsigned long drawable, int interval) {
+    GLXImpl::SwapIntervalEXT(dpy, drawable, interval);
+}
+
+MOBILEGL_GLX_API int glXSwapIntervalMESA(unsigned int interval) {
+    return GLXImpl::SwapIntervalMESA(interval);
+}
+
+MOBILEGL_GLX_API int glXGetSwapIntervalMESA() {
+    return GLXImpl::GetSwapIntervalMESA();
+}
+
+MOBILEGL_GLX_API int glXSwapIntervalSGI(int interval) {
+    return GLXImpl::SwapIntervalSGI(interval);
+}
+
+// Legacy entry points some loaders probe for; harmless no-op stubs.
+MOBILEGL_GLX_API void glXCopyContext(Display*, void*, void*, unsigned long) {
+    MGLOG_W_ONCE("glx: glXCopyContext is not supported");
+}
+
+MOBILEGL_GLX_API unsigned long glXCreateGLXPixmap(Display*, void*, unsigned long) {
+    MGLOG_W_ONCE("glx: glXCreateGLXPixmap is not supported");
+    return 0;
+}
+
+MOBILEGL_GLX_API void glXDestroyGLXPixmap(Display*, unsigned long) {}
+
+MOBILEGL_GLX_API unsigned long glXCreatePixmap(Display*, void*, unsigned long, const int*) {
+    MGLOG_W_ONCE("glx: glXCreatePixmap is not supported");
+    return 0;
+}
+
+MOBILEGL_GLX_API void glXDestroyPixmap(Display*, unsigned long) {}
+
+MOBILEGL_GLX_API unsigned long glXCreatePbuffer(Display*, void*, const int*) {
+    MGLOG_W_ONCE("glx: glXCreatePbuffer is not supported");
+    return 0;
+}
+
+MOBILEGL_GLX_API void glXDestroyPbuffer(Display*, unsigned long) {}
+
+MOBILEGL_GLX_API void glXUseXFont(unsigned long, int, int, int) {
+    MGLOG_W_ONCE("glx: glXUseXFont is not supported");
+}
+
+MOBILEGL_GLX_API void glXSelectEvent(Display*, unsigned long, unsigned long) {}
+
+MOBILEGL_GLX_API void glXGetSelectedEvent(Display*, unsigned long, unsigned long* eventMask) {
+    if (eventMask) {
+        *eventMask = 0;
+    }
+}
+
+namespace MobileGL::MG_Impl::GLXImpl {
+    namespace {
+        struct GLXEntryPoint {
+            const char* Name;
+            void* Proc;
+        };
+
+        const GLXEntryPoint kGLXEntryPoints[] = {
+            {"glXChooseFBConfig", reinterpret_cast<void*>(glXChooseFBConfig)},
+            {"glXChooseVisual", reinterpret_cast<void*>(glXChooseVisual)},
+            {"glXCopyContext", reinterpret_cast<void*>(glXCopyContext)},
+            {"glXCreateContext", reinterpret_cast<void*>(glXCreateContext)},
+            {"glXCreateContextAttribsARB", reinterpret_cast<void*>(glXCreateContextAttribsARB)},
+            {"glXCreateGLXPixmap", reinterpret_cast<void*>(glXCreateGLXPixmap)},
+            {"glXCreateNewContext", reinterpret_cast<void*>(glXCreateNewContext)},
+            {"glXCreatePbuffer", reinterpret_cast<void*>(glXCreatePbuffer)},
+            {"glXCreatePixmap", reinterpret_cast<void*>(glXCreatePixmap)},
+            {"glXCreateWindow", reinterpret_cast<void*>(glXCreateWindow)},
+            {"glXDestroyContext", reinterpret_cast<void*>(glXDestroyContext)},
+            {"glXDestroyGLXPixmap", reinterpret_cast<void*>(glXDestroyGLXPixmap)},
+            {"glXDestroyPbuffer", reinterpret_cast<void*>(glXDestroyPbuffer)},
+            {"glXDestroyPixmap", reinterpret_cast<void*>(glXDestroyPixmap)},
+            {"glXDestroyWindow", reinterpret_cast<void*>(glXDestroyWindow)},
+            {"glXGetClientString", reinterpret_cast<void*>(glXGetClientString)},
+            {"glXGetConfig", reinterpret_cast<void*>(glXGetConfig)},
+            {"glXGetCurrentContext", reinterpret_cast<void*>(glXGetCurrentContext)},
+            {"glXGetCurrentDisplay", reinterpret_cast<void*>(glXGetCurrentDisplay)},
+            {"glXGetCurrentDrawable", reinterpret_cast<void*>(glXGetCurrentDrawable)},
+            {"glXGetCurrentReadDrawable", reinterpret_cast<void*>(glXGetCurrentReadDrawable)},
+            {"glXGetFBConfigAttrib", reinterpret_cast<void*>(glXGetFBConfigAttrib)},
+            {"glXGetFBConfigs", reinterpret_cast<void*>(glXGetFBConfigs)},
+            {"glXGetProcAddress", reinterpret_cast<void*>(glXGetProcAddress)},
+            {"glXGetProcAddressARB", reinterpret_cast<void*>(glXGetProcAddressARB)},
+            {"glXGetSelectedEvent", reinterpret_cast<void*>(glXGetSelectedEvent)},
+            {"glXGetSwapIntervalMESA", reinterpret_cast<void*>(glXGetSwapIntervalMESA)},
+            {"glXGetVisualFromFBConfig", reinterpret_cast<void*>(glXGetVisualFromFBConfig)},
+            {"glXIsDirect", reinterpret_cast<void*>(glXIsDirect)},
+            {"glXMakeContextCurrent", reinterpret_cast<void*>(glXMakeContextCurrent)},
+            {"glXMakeCurrent", reinterpret_cast<void*>(glXMakeCurrent)},
+            {"glXQueryContext", reinterpret_cast<void*>(glXQueryContext)},
+            {"glXQueryDrawable", reinterpret_cast<void*>(glXQueryDrawable)},
+            {"glXQueryExtension", reinterpret_cast<void*>(glXQueryExtension)},
+            {"glXQueryExtensionsString", reinterpret_cast<void*>(glXQueryExtensionsString)},
+            {"glXQueryServerString", reinterpret_cast<void*>(glXQueryServerString)},
+            {"glXQueryVersion", reinterpret_cast<void*>(glXQueryVersion)},
+            {"glXSelectEvent", reinterpret_cast<void*>(glXSelectEvent)},
+            {"glXSwapBuffers", reinterpret_cast<void*>(glXSwapBuffers)},
+            {"glXSwapIntervalEXT", reinterpret_cast<void*>(glXSwapIntervalEXT)},
+            {"glXSwapIntervalMESA", reinterpret_cast<void*>(glXSwapIntervalMESA)},
+            {"glXSwapIntervalSGI", reinterpret_cast<void*>(glXSwapIntervalSGI)},
+            {"glXUseXFont", reinterpret_cast<void*>(glXUseXFont)},
+            {"glXWaitGL", reinterpret_cast<void*>(glXWaitGL)},
+            {"glXWaitX", reinterpret_cast<void*>(glXWaitX)},
+        };
+    } // namespace
+
+    void* GetGLXEntryPoint(const char* name) {
+        for (const auto& entry : kGLXEntryPoints) {
+            if (std::strcmp(entry.Name, name) == 0) {
+                return entry.Proc;
+            }
+        }
+        return nullptr;
+    }
+} // namespace MobileGL::MG_Impl::GLXImpl
+
+#endif // __linux__ && !__ANDROID__

MobileGL/MG_Impl/GLXImpl/GLXImpl.h

@@ -0,0 +1,69 @@
+// MobileGL - MobileGL/MG_Impl/GLXImpl/GLXImpl.h
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#pragma once
+#include <Includes.h>
+
+#if defined(__linux__) && !defined(__ANDROID__)
+
+namespace MobileGL::MG_Impl::GLXImpl {
+    // GLX layered on MobileGL's own EGL, mirroring WGLImpl/CGLImpl. Handles are
+    // opaque to callers; XVisualInfo crosses the ABI as void* so this header
+    // needs no Xlib includes (Includes.h forward-declares Display/XID/Window).
+    using GLXFBConfigHandle = void*;
+    using GLXContextHandle = void*;
+    using GLXDrawableHandle = unsigned long; // XID
+
+    int QueryExtension(Display* dpy, int* errorBase, int* eventBase);
+    int QueryVersion(Display* dpy, int* major, int* minor);
+    const char* QueryExtensionsString(Display* dpy, int screen);
+    const char* GetClientString(Display* dpy, int name);
+    const char* QueryServerString(Display* dpy, int screen, int name);
+
+    GLXFBConfigHandle* GetFBConfigs(Display* dpy, int screen, int* nelements);
+    GLXFBConfigHandle* ChooseFBConfig(Display* dpy, int screen, const int* attribList, int* nelements);
+    int GetFBConfigAttrib(Display* dpy, GLXFBConfigHandle config, int attribute, int* value);
+    void* GetVisualFromFBConfig(Display* dpy, GLXFBConfigHandle config);
+    void* ChooseVisual(Display* dpy, int screen, int* attribList);
+    int GetConfig(Display* dpy, void* visualInfo, int attribute, int* value);
+
+    GLXContextHandle CreateContext(Display* dpy, void* visualInfo, GLXContextHandle share, int direct);
+    GLXContextHandle CreateNewContext(Display* dpy, GLXFBConfigHandle config, int renderType,
+                                      GLXContextHandle share, int direct);
+    GLXContextHandle CreateContextAttribsARB(Display* dpy, GLXFBConfigHandle config, GLXContextHandle share,
+                                             int direct, const int* attribList);
+    void DestroyContext(Display* dpy, GLXContextHandle context);
+    int MakeCurrent(Display* dpy, GLXDrawableHandle drawable, GLXContextHandle context);
+    int MakeContextCurrent(Display* dpy, GLXDrawableHandle draw, GLXDrawableHandle read,
+                           GLXContextHandle context);
+    void SwapBuffers(Display* dpy, GLXDrawableHandle drawable);
+
+    GLXDrawableHandle CreateWindow(Display* dpy, GLXFBConfigHandle config, GLXDrawableHandle window,
+                                   const int* attribList);
+    void DestroyWindow(Display* dpy, GLXDrawableHandle window);
+
+    GLXContextHandle GetCurrentContext();
+    GLXDrawableHandle GetCurrentDrawable();
+    GLXDrawableHandle GetCurrentReadDrawable();
+    Display* GetCurrentDisplay();
+    int IsDirect(Display* dpy, GLXContextHandle context);
+    void WaitGL();
+    void WaitX();
+    int QueryContext(Display* dpy, GLXContextHandle context, int attribute, int* value);
+    void QueryDrawable(Display* dpy, GLXDrawableHandle drawable, int attribute, unsigned int* value);
+
+    void SwapIntervalEXT(Display* dpy, GLXDrawableHandle drawable, int interval);
+    int SwapIntervalMESA(unsigned int interval);
+    int GetSwapIntervalMESA();
+    int SwapIntervalSGI(int interval);
+
+    // Name -> exported glX entry point (table lives with the exports).
+    void* GetGLXEntryPoint(const char* name);
+} // namespace MobileGL::MG_Impl::GLXImpl
+
+#endif // __linux__ && !__ANDROID__

MobileGL/MG_Impl/GLXImpl/LookUp/LookUp.cpp

@@ -8,14 +8,30 @@
 
 #include "LookUp.h"
 
-namespace MG_Impl::GLXImpl {
+#if defined(__linux__) && !defined(__ANDROID__)
-    // TODO: implement complete GLX functionality
+#include "../GLXImpl.h"
+#endif
 
+namespace MG_Impl::GLXImpl {
     void* GetProcAddress(const char* name) {
+        if (!name) {
+            return nullptr;
+        }
         MGLOG_D("glXGetProcAddress(\"%s\")", name);
+#if defined(__linux__) && !defined(__ANDROID__)
+        if (name[0] == 'g' && name[1] == 'l' && name[2] == 'X') {
+            // glX entry points resolve from the GLX layer's own table; GL/EGL
+            // names fall through to the shared resolver below.
+            void* proc = MobileGL::MG_Impl::GLXImpl::GetGLXEntryPoint(name);
+            if (!proc) {
+                MGLOG_D("glXGetProcAddress: unknown glX entry point %s", name);
+            }
+            return proc;
+        }
+#endif
         void* proc = MobileGL::MG_Impl::GetProcAddress(name);
         if (!proc) {
-            MGLOG_W("Failed to get function: %s", (const char*)name);
+            MGLOG_D("Failed to get function: %s", (const char*)name);
             return nullptr;
         }
 

MobileGL/MG_Impl/GetProcAddress.cpp

@@ -1403,7 +1403,7 @@ namespace MobileGL::MG_Impl {
         GETPROC(glFramebufferTextureMultiviewOVR, name);
         // GETPROC(glNamedFramebufferTextureMultiviewOVR, name);
 
-        MGLOG_W("GetProcAddress(%s) = nullptr!", name);
+        MGLOG_D("GetProcAddress(%s) = nullptr!", name);
         return nullptr;
     }
 } // namespace MobileGL::MG_Impl

MobileGL/MG_Impl/NSOpenGLImpl/NSOpenGLImpl.cpp

@@ -269,7 +269,7 @@ namespace MobileGL::MG_Impl::NSOpenGLImpl {
             }
             id metalLayerClass = reinterpret_cast<id>(objc_getClass("CAMetalLayer"));
             if (!metalLayerClass) {
-                MGLOG_E("NSOpenGLImpl: CAMetalLayer class not found");
+                MGLOG_E_ONCE("NSOpenGLImpl: CAMetalLayer class not found");
                 return nil;
             }
 
@@ -310,7 +310,7 @@ namespace MobileGL::MG_Impl::NSOpenGLImpl {
                                                        static_cast<GLint>(geometry.DrawableSize.width),
                                                        static_cast<GLint>(geometry.DrawableSize.height));
             if (error != kCGLNoError) {
-                MGLOG_E("NSOpenGLImpl: failed to attach drawable: %s", CGLImpl::ErrorString(error));
+                MGLOG_E_ONCE("NSOpenGLImpl: failed to attach drawable: %s", CGLImpl::ErrorString(error));
             }
         }
 
@@ -325,7 +325,7 @@ namespace MobileGL::MG_Impl::NSOpenGLImpl {
             }
             const auto error = CGLImpl::SetCurrentContext(context);
             if (error != kCGLNoError) {
-                MGLOG_E("NSOpenGLImpl: makeCurrentContext failed: %s", CGLImpl::ErrorString(error));
+                MGLOG_E_ONCE("NSOpenGLImpl: makeCurrentContext failed: %s", CGLImpl::ErrorString(error));
             }
         }
 
@@ -345,7 +345,7 @@ namespace MobileGL::MG_Impl::NSOpenGLImpl {
             }
             const auto error = CGLImpl::FlushDrawable(context);
             if (error != kCGLNoError) {
-                MGLOG_E("NSOpenGLImpl: flushBuffer failed: %s", CGLImpl::ErrorString(error));
+                MGLOG_E_ONCE("NSOpenGLImpl: flushBuffer failed: %s", CGLImpl::ErrorString(error));
             }
         }
 
@@ -377,7 +377,7 @@ namespace MobileGL::MG_Impl::NSOpenGLImpl {
                                                        static_cast<GLint>(geometry.DrawableSize.width),
                                                        static_cast<GLint>(geometry.DrawableSize.height));
             if (error != kCGLNoError) {
-                MGLOG_E("NSOpenGLImpl: update failed to attach drawable: %s", CGLImpl::ErrorString(error));
+                MGLOG_E_ONCE("NSOpenGLImpl: update failed to attach drawable: %s", CGLImpl::ErrorString(error));
                 return;
             }
             CGLImpl::UpdateContext(context);
@@ -421,7 +421,7 @@ namespace MobileGL::MG_Impl::NSOpenGLImpl {
             SEL selector = sel_registerName(selectorName);
             Method method = class_getInstanceMethod(cls, selector);
             if (!method) {
-                MGLOG_W("NSOpenGLImpl: missing instance method %s", selectorName);
+                MGLOG_W_ONCE("NSOpenGLImpl: missing instance method %s", selectorName);
                 return;
             }
             if (original) {
@@ -434,7 +434,7 @@ namespace MobileGL::MG_Impl::NSOpenGLImpl {
             SEL selector = sel_registerName(selectorName);
             Method method = class_getClassMethod(cls, selector);
             if (!method) {
-                MGLOG_W("NSOpenGLImpl: missing class method %s", selectorName);
+                MGLOG_W_ONCE("NSOpenGLImpl: missing class method %s", selectorName);
                 return;
             }
             method_setImplementation(method, replacement);
@@ -444,7 +444,7 @@ namespace MobileGL::MG_Impl::NSOpenGLImpl {
             Class pixelFormatClass = objc_getClass("NSOpenGLPixelFormat");
             Class contextClass = objc_getClass("NSOpenGLContext");
             if (!pixelFormatClass || !contextClass) {
-                MGLOG_W("NSOpenGLImpl: NSOpenGL classes are not loaded; hooks not installed");
+                MGLOG_W_ONCE("NSOpenGLImpl: NSOpenGL classes are not loaded; hooks not installed");
                 return false;
             }
 

MobileGL/MG_Impl/WGLImpl/Exporting/Definitions.cpp

@@ -56,7 +56,7 @@ extern "C" HGLRC WINAPI wglCreateLayerContext(HDC hdc, int iLayerPlane) {
 }
 
 extern "C" BOOL WINAPI wglCopyContext(HGLRC, HGLRC, UINT) {
-    MGLOG_W("wglCopyContext is not supported");
+    MGLOG_W_ONCE("wglCopyContext is not supported");
     SetLastError(ERROR_NOT_SUPPORTED);
     return FALSE;
 }
@@ -132,24 +132,24 @@ extern "C" DWORD WINAPI wglSwapMultipleBuffers(UINT n, CONST WGLSWAP* ps) {
 // ---- Font rendering (legacy immediate-mode feature; not supported) ----
 
 extern "C" BOOL WINAPI wglUseFontBitmapsA(HDC, DWORD, DWORD, DWORD) {
-    MGLOG_W("wglUseFontBitmapsA is not supported");
+    MGLOG_W_ONCE("wglUseFontBitmapsA is not supported");
     return FALSE;
 }
 
 extern "C" BOOL WINAPI wglUseFontBitmapsW(HDC, DWORD, DWORD, DWORD) {
-    MGLOG_W("wglUseFontBitmapsW is not supported");
+    MGLOG_W_ONCE("wglUseFontBitmapsW is not supported");
     return FALSE;
 }
 
 extern "C" BOOL WINAPI wglUseFontOutlinesA(HDC, DWORD, DWORD, DWORD, FLOAT, FLOAT, int,
                                                  LPGLYPHMETRICSFLOAT) {
-    MGLOG_W("wglUseFontOutlinesA is not supported");
+    MGLOG_W_ONCE("wglUseFontOutlinesA is not supported");
     return FALSE;
 }
 
 extern "C" BOOL WINAPI wglUseFontOutlinesW(HDC, DWORD, DWORD, DWORD, FLOAT, FLOAT, int,
                                                  LPGLYPHMETRICSFLOAT) {
-    MGLOG_W("wglUseFontOutlinesW is not supported");
+    MGLOG_W_ONCE("wglUseFontOutlinesW is not supported");
     return FALSE;
 }
 

MobileGL/MG_Impl/WGLImpl/WGLImpl.cpp

@@ -215,7 +215,7 @@ namespace MobileGL::MG_Impl::WGLImpl {
             Uint32 width = 0;
             Uint32 height = 0;
             if (!QueryClientSize(hwnd, width, height)) {
-                MGLOG_E("wgl: GetClientRect failed for HWND %p", hwnd);
+                MGLOG_E_ONCE("wgl: GetClientRect failed for HWND %p", hwnd);
                 return nullptr;
             }
 
@@ -227,7 +227,7 @@ namespace MobileGL::MG_Impl::WGLImpl {
             EGLSurface surface =
                 EGLImpl::CreatePlatformWindowSurface(context.Display, context.Config, hwnd, attribs);
             if (surface == EGL_NO_SURFACE) {
-                MGLOG_E("wgl: failed to create window surface for HWND %p (%ux%u)", hwnd, width, height);
+                MGLOG_E_ONCE("wgl: failed to create window surface for HWND %p (%ux%u)", hwnd, width, height);
                 return nullptr;
             }
 
@@ -244,7 +244,7 @@ namespace MobileGL::MG_Impl::WGLImpl {
             const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
             EGLDisplay display = EnsureDisplay();
             if (display == EGL_NO_DISPLAY) {
-                MGLOG_E("wgl: no EGL display");
+                MGLOG_E_ONCE("wgl: no EGL display");
                 return nullptr;
             }
             EGLImpl::BindAPI(EGL_OPENGL_API);
@@ -275,13 +275,13 @@ namespace MobileGL::MG_Impl::WGLImpl {
             EGLConfig config = nullptr;
             EGLint configCount = 0;
             if (!EGLImpl::ChooseConfig(display, configAttribs, &config, 1, &configCount) || configCount <= 0) {
-                MGLOG_E("wgl: eglChooseConfig failed");
+                MGLOG_E_ONCE("wgl: eglChooseConfig failed");
                 return nullptr;
             }
 
             EGLContext eglContext = EGLImpl::CreateContext(display, config, shareContext, contextAttribs);
             if (eglContext == EGL_NO_CONTEXT) {
-                MGLOG_E("wgl: eglCreateContext failed");
+                MGLOG_E_ONCE("wgl: eglCreateContext failed");
                 return nullptr;
             }
 
@@ -612,7 +612,7 @@ namespace MobileGL::MG_Impl::WGLImpl {
         auto& surfaces = WindowSurfaces();
         auto it = surfaces.find(hwnd);
         if (it == surfaces.end()) {
-            MGLOG_W("wglSwapBuffers: no surface for HWND %p", hwnd);
+            MGLOG_W_ONCE("wglSwapBuffers: no surface for HWND %p", hwnd);
             return FALSE;
         }
         SyncSurfaceSize(hwnd, it->second);
@@ -685,7 +685,7 @@ namespace MobileGL::MG_Impl::WGLImpl {
         }
 
         if (!EGLImpl::MakeCurrent(object->Display, surface->Surface, surface->Surface, object->Context)) {
-            MGLOG_E("wglMakeCurrent: eglMakeCurrent failed (hdc=%p, hglrc=%p)", hdc, hglrc);
+            MGLOG_E_ONCE("wglMakeCurrent: eglMakeCurrent failed (hdc=%p, hglrc=%p)", hdc, hglrc);
             return FALSE;
         }
         t_current = {hdc, hglrc};

MobileGL/MG_IntegrationTest/CMakeLists.txt

@@ -0,0 +1,341 @@
+cmake_minimum_required(VERSION 3.24)
+
+# MobileGL headless GPU integration tests.
+#
+# These are not unit tests: each scenario brings up a real EGL context on a
+# pbuffer, renders real frames through a real backend and asserts on
+# glReadPixels output. They need a GPU, so the module is OFF by default
+# (MOBILEGL_BUILD_INTEGRATION_TEST) and every scenario skips cleanly - never
+# fails, never hangs - on a machine without one. "Cleanly" is not a hope: the
+# harness runs the whole bring-up in a forked child first, because MobileGL
+# ABORTS rather than returning an error on an unusable platform (HeadlessGL.cpp).
+#
+# A clean skip is also indistinguishable from a pass, so set
+# MOBILEGL_ITEST_REQUIRE_GPU wherever the machine is supposed to have a GPU.
+#
+# Backend selection is latched at initialization from MOBILEGL_BACKEND_TYPE, so
+# one process is one backend: the same binary is registered twice, once per
+# backend, under the `integration-gpu` label.
+
+message(STATUS "Generating build files for MobileGL Integration Test...")
+
+set(CMAKE_CXX_STANDARD 23)
+set(CMAKE_CXX_STANDARD_REQUIRED ON)
+
+set(MGL_ITEST_ROOT ${CMAKE_CURRENT_LIST_DIR}/../..)
+
+# Desktop links the static implementation directly. Android runs the same
+# executable from adb shell and links the shipping shared library instead.
+if (ANDROID)
+    set(MGL_ITEST_MOBILEGL_TARGET MobileGL)
+elseif (TARGET MobileGL_s)
+    set(MGL_ITEST_MOBILEGL_TARGET MobileGL_s)
+else()
+    message(STATUS "No MobileGL library target is available; skipping the integration test module")
+    return()
+endif()
+
+# MG_Test already pulls googletest in when MOBILEGL_BUILD_TEST is ON. Stand on
+# our own feet when it is not, so this module can be built by itself.
+if (NOT TARGET GTest::gtest)
+    include(FetchContent)
+    FetchContent_Declare(
+        googletest
+        GIT_REPOSITORY https://github.com/google/googletest.git
+        GIT_TAG        v1.17.0
+    )
+    set(gtest_force_shared_crt ON CACHE BOOL "" FORCE)
+    FetchContent_MakeAvailable(googletest)
+endif()
+
+add_executable(MobileGLIntegrationTest
+    Main.cpp
+    Harness/HeadlessGL.cpp
+    Scenarios/OrientationScenario.cpp
+    Scenarios/CrossFrameBufferScenario.cpp
+    Scenarios/ResidentIndexScenario.cpp
+    Scenarios/MultiDrawScenario.cpp
+    Scenarios/DrawParametersScenario.cpp
+    Scenarios/AsyncCompileScenario.cpp
+    Scenarios/XfbAfterClipDistanceScenario.cpp
+    Scenarios/ThreeChannelAttachmentScenario.cpp
+    Scenarios/PipelineFailureScenario.cpp
+    Scenarios/AdvertisedLimitsScenario.cpp
+    Scenarios/PixelStoreSweepScenario.cpp
+    Scenarios/FragCoordOriginScenario.cpp
+    Scenarios/ClearThenReadPixelsScenario.cpp
+    Scenarios/DepthStencilReadbackScenario.cpp
+    Scenarios/DepthStencilReadbackMatrixScenario.cpp
+    Scenarios/DepthStencilReadbackAttachmentShapeScenario.cpp
+    Scenarios/ClipDistanceScenario.cpp
+    Scenarios/ViewportArrayScenario.cpp
+    Scenarios/SsboArrayLengthScenario.cpp
+    Scenarios/DoublePrecisionScenario.cpp
+    Scenarios/UniformInitializerScenario.cpp
+    Scenarios/SwizzleAccessRoutineScenario.cpp
+    Scenarios/IterationRPFirstReductionScenario.cpp
+    Scenarios/IterationRPProgram203Scenario.cpp
+    Scenarios/IterationRPScratchFixScenario.cpp
+    Scenarios/ProgramPipelineScenario.cpp
+    Scenarios/ImageLoadStoreSsoScenario.cpp
+    Scenarios/ImageTargetKindScenario.cpp
+    Scenarios/ImageFormatQualifierScenario.cpp
+    Scenarios/SsboDeclarationFormScenario.cpp
+    Scenarios/Glsl420DeclarationScenario.cpp
+    Scenarios/FragmentOutputArrayIndexScenario.cpp
+    Scenarios/BufferTextureScenario.cpp
+    Scenarios/VertexAttribBindingScenario.cpp
+    Scenarios/XfbCaptureBufferReuseScenario.cpp
+    Scenarios/VertexArrayEnableDisableScenario.cpp
+    Scenarios/CopyImageLevelRangeScenario.cpp
+    Scenarios/CopyImageLayeredScenario.cpp
+    Scenarios/LayeredAttachmentBarrierScenario.cpp
+)
+
+target_include_directories(MobileGLIntegrationTest PRIVATE
+    ${MGL_ITEST_ROOT}/include
+    ${MGL_ITEST_ROOT}/MobileGL
+)
+
+# gtest, not gtest_main: Main.cpp installs the harness banner itself.
+target_link_libraries(MobileGLIntegrationTest PRIVATE
+    GTest::gtest
+    ${MGL_ITEST_MOBILEGL_TARGET}
+)
+
+if (ANDROID)
+    find_library(MGL_ITEST_ANDROID_LIBRARY android REQUIRED)
+    find_library(MGL_ITEST_LOG_LIBRARY log REQUIRED)
+    find_library(MGL_ITEST_MEDIANDK_LIBRARY mediandk REQUIRED)
+    target_link_libraries(MobileGLIntegrationTest PRIVATE
+        ${MGL_ITEST_ANDROID_LIBRARY}
+        ${MGL_ITEST_LOG_LIBRARY}
+        ${MGL_ITEST_MEDIANDK_LIBRARY}
+    )
+endif()
+
+if (MSVC)
+    # Same reason as MG_Test/Backend/DirectVulkan: the GLES headers declare gl*
+    # as dllimport on Windows, so the in-library GL entry-point definitions only
+    # resolve if the whole static library is part of the link.
+    target_link_options(MobileGLIntegrationTest PRIVATE /WHOLEARCHIVE:MobileGL_s)
+endif()
+target_compile_definitions(MobileGLIntegrationTest PRIVATE -DNOMINMAX)
+
+if (ANDROID)
+    return()
+endif()
+
+# --- ctest wiring --------------------------------------------------------
+# A bare libEGL on a glvnd box resolves to whatever vendor comes first, which is
+# usually Mesa/llvmpipe - a software rasteriser silently replacing the GPU under
+# a GPU test. Pin the vendor/ICD json the same way MG_Benchmark's
+# run_driver_bench.sh does.
+#
+# Leaving these empty is not a neutral default, it is the failure mode: an
+# unpinned libEGL lands on llvmpipe and the suite goes green having tested a
+# software rasteriser. So they are DETECTED here rather than defaulted to empty,
+# and an empty result is a loud warning.
+#
+# mgl_itest_find_driver_json(<outVar> <description> <glob> [<glob>...])
+#   Picks the first json a real hardware vendor owns, in preference order, and
+#   never picks a software rasteriser (llvmpipe / lavapipe / swrast) - landing on
+#   one of those silently is the exact accident this pinning exists to prevent.
+function(mgl_itest_find_driver_json outVar)
+    set(candidates "")
+    foreach(pattern IN LISTS ARGN)
+        file(GLOB matches "${pattern}")
+        list(APPEND candidates ${matches})
+    endforeach()
+    list(SORT candidates)
+    # Vendors ship an i686 json beside the x86_64 one and it sorts first. Pinning
+    # the wrong word size is worse than not pinning at all - the loader finds no
+    # driver and the whole suite skips - so drop the mismatched ones outright.
+    if (CMAKE_SIZEOF_VOID_P EQUAL 8)
+        list(FILTER candidates EXCLUDE REGEX "i686|i386")
+    else()
+        list(FILTER candidates EXCLUDE REGEX "x86_64|aarch64")
+    endif()
+    set(software "")
+    foreach(vendor IN ITEMS nvidia amdgpu amd radeon intel_hasvk intel broadcom freedreno panfrost)
+        foreach(candidate IN LISTS candidates)
+            get_filename_component(leaf "${candidate}" NAME)
+            string(TOLOWER "${leaf}" leaf)
+            if (leaf MATCHES "${vendor}")
+                set(${outVar} "${candidate}" PARENT_SCOPE)
+                return()
+            endif()
+        endforeach()
+    endforeach()
+    # Nothing recognised as hardware. Report the first non-software entry if there
+    # is one; otherwise report nothing, so the warning below fires.
+    foreach(candidate IN LISTS candidates)
+        get_filename_component(leaf "${candidate}" NAME)
+        string(TOLOWER "${leaf}" leaf)
+        if (NOT leaf MATCHES "lvp|llvmpipe|lavapipe|swrast|softpipe")
+            set(${outVar} "${candidate}" PARENT_SCOPE)
+            return()
+        endif()
+        set(software "${candidate}")
+    endforeach()
+    set(${outVar} "" PARENT_SCOPE)
+endfunction()
+
+set(MGL_ITEST_DETECTED_EGL_VENDOR "")
+set(MGL_ITEST_DETECTED_VK_ICD "")
+if (UNIX AND NOT APPLE AND NOT ANDROID)
+    mgl_itest_find_driver_json(MGL_ITEST_DETECTED_EGL_VENDOR
+        "/usr/share/glvnd/egl_vendor.d/*.json"
+        "/etc/glvnd/egl_vendor.d/*.json")
+    mgl_itest_find_driver_json(MGL_ITEST_DETECTED_VK_ICD
+        "/usr/share/vulkan/icd.d/*.json"
+        "/etc/vulkan/icd.d/*.json")
+endif()
+
+set(MOBILEGL_ITEST_EGL_VENDOR "${MGL_ITEST_DETECTED_EGL_VENDOR}" CACHE FILEPATH
+    "glvnd EGL vendor json to pin for the integration tests (empty: leave the loader alone)")
+set(MOBILEGL_ITEST_VK_ICD "${MGL_ITEST_DETECTED_VK_ICD}" CACHE FILEPATH
+    "Vulkan ICD json to pin for the DirectVulkan integration tests (empty: leave the loader alone)")
+
+if (MOBILEGL_ITEST_EGL_VENDOR)
+    message(STATUS "Integration tests: pinning EGL vendor ${MOBILEGL_ITEST_EGL_VENDOR}")
+else()
+    message(WARNING
+        "Integration tests: no EGL vendor json found or configured (MOBILEGL_ITEST_EGL_VENDOR is empty). "
+        "An unpinned libEGL on a glvnd system resolves to whichever vendor comes first, which is usually "
+        "Mesa/llvmpipe - the scenarios would then go green against a software rasteriser instead of the GPU. "
+        "Set -DMOBILEGL_ITEST_EGL_VENDOR=/usr/share/glvnd/egl_vendor.d/<vendor>.json.")
+endif()
+if (MOBILEGL_ITEST_VK_ICD)
+    message(STATUS "Integration tests: pinning Vulkan ICD ${MOBILEGL_ITEST_VK_ICD}")
+else()
+    message(WARNING
+        "Integration tests: no Vulkan ICD json found or configured (MOBILEGL_ITEST_VK_ICD is empty). "
+        "DirectVulkan would then load whichever ICD the loader enumerates first, quite possibly lavapipe. "
+        "Set -DMOBILEGL_ITEST_VK_ICD=/usr/share/vulkan/icd.d/<vendor>.json.")
+endif()
+
+# Turns "no usable GPU" from a clean skip into a failure - see ScenarioFixture.h.
+# Without it the integration-gpu label is unfalsifiable: a run that skipped every
+# scenario and a run that passed every scenario are the same green in ctest.
+option(MOBILEGL_ITEST_REQUIRE_GPU
+    "Fail (rather than skip) the integration scenarios when the headless harness is unusable" OFF)
+
+# No EGL_PLATFORM knob here on purpose. The harness pins EGL_PLATFORM=surfaceless
+# itself before its first EGL call (HeadlessGL.cpp, EnsureHeadlessPlatform) so a
+# developer's machine and a CI runner take the SAME path whether or not a window
+# system happens to be running. This used to inject "x11", which is how the lane
+# came up green on a workstation with WSLg and died on a runner with no X server.
+#
+# A build-system knob would not just be redundant, it would be a trap: `set(...
+# CACHE ...)` does not rewrite an existing cache, so every build directory
+# configured before this change would keep injecting EGL_PLATFORM=x11 and go on
+# binding to a window system - silently, and only on the machines that have one.
+# Someone reproducing a platform-specific bug sets EGL_PLATFORM in their own
+# environment, which the harness still honours.
+
+set(MGL_ITEST_COMMON_ENV "")
+if (MOBILEGL_ITEST_EGL_VENDOR)
+    list(APPEND MGL_ITEST_COMMON_ENV "__EGL_VENDOR_LIBRARY_FILENAMES=${MOBILEGL_ITEST_EGL_VENDOR}")
+endif()
+unset(MOBILEGL_ITEST_EGL_PLATFORM CACHE) # see above: an old cache must not resurrect x11
+if (MOBILEGL_ITEST_REQUIRE_GPU)
+    list(APPEND MGL_ITEST_COMMON_ENV "MOBILEGL_ITEST_REQUIRE_GPU=1")
+endif()
+
+set(MGL_ITEST_VULKAN_ENV ${MGL_ITEST_COMMON_ENV})
+if (MOBILEGL_ITEST_VK_ICD)
+    list(APPEND MGL_ITEST_VULKAN_ENV "VK_ICD_FILENAMES=${MOBILEGL_ITEST_VK_ICD}")
+endif()
+
+# The ENVIRONMENT test property is itself a `;`-list, and gtest_discover_tests
+# forwards PROPERTIES as a flat list - so a plain `;`-joined value arrives as
+# four separate arguments and everything after the first is silently read as
+# another property name. Escaping the separators keeps the whole thing one list
+# element until set_tests_properties expands it back. Without this only
+# MOBILEGL_BACKEND_TYPE reaches the test and the vendor/ICD pinning is lost.
+function(mgl_itest_join_environment outVar)
+    set(joined "")
+    foreach(entry IN LISTS ARGN)
+        if (joined)
+            string(APPEND joined "\\;${entry}")
+        else()
+            set(joined "${entry}")
+        endif()
+    endforeach()
+    set(${outVar} "${joined}" PARENT_SCOPE)
+endfunction()
+
+mgl_itest_join_environment(MGL_ITEST_GLES_ENVIRONMENT
+    "MOBILEGL_BACKEND_TYPE=DirectGLES" ${MGL_ITEST_COMMON_ENV})
+mgl_itest_join_environment(MGL_ITEST_VULKAN_ENVIRONMENT
+    "MOBILEGL_BACKEND_TYPE=DirectVulkan" ${MGL_ITEST_VULKAN_ENV})
+mgl_itest_join_environment(MGL_ITEST_VULKAN_ASYNC_ENVIRONMENT
+    "MOBILEGL_BACKEND_TYPE=DirectVulkan" "MOBILEGL_ASYNC_SHADER_COMPILE=1" ${MGL_ITEST_VULKAN_ENV})
+mgl_itest_join_environment(MGL_ITEST_GLES_FORCED_DS_ENVIRONMENT
+    "MOBILEGL_BACKEND_TYPE=DirectGLES" "MOBILEGL_ESPRYT_FORCE_DS_READBACK_EMULATION=1" ${MGL_ITEST_COMMON_ENV})
+
+# TIMEOUT on every entry: a GPU test that wedges must fail the run, not hang it.
+set(MGL_ITEST_TIMEOUT 120)
+
+include(GoogleTest)
+
+# Discovery runs `--gtest_list_tests`, which does not construct the harness and
+# so needs no GPU. One registration per backend; TEST_PREFIX keeps the two sets
+# of ctest names apart.
+gtest_discover_tests(MobileGLIntegrationTest
+    TEST_PREFIX "DirectGLES."
+    DISCOVERY_TIMEOUT 30
+    PROPERTIES
+        LABELS integration-gpu
+        TIMEOUT ${MGL_ITEST_TIMEOUT}
+        ENVIRONMENT "${MGL_ITEST_GLES_ENVIRONMENT}"
+)
+
+gtest_discover_tests(MobileGLIntegrationTest
+    TEST_PREFIX "DirectVulkan."
+    DISCOVERY_TIMEOUT 30
+    PROPERTIES
+        LABELS integration-gpu
+        TIMEOUT ${MGL_ITEST_TIMEOUT}
+        ENVIRONMENT "${MGL_ITEST_VULKAN_ENVIRONMENT}"
+)
+
+# A third registration, of ONE scenario, with asynchronous shader compilation
+# pinned on. Not a second code path in the renderer: a second ALLOCATION pattern.
+# The async pipeline's job objects change which of the freed blocks the capture
+# phase is handed, and that is what decides whether the destroyed-VAO address is
+# reached at all - on the ablated (pre-fix) tree async=1 reproduced 3 runs out of
+# 3 where the ambient default reproduced 2 of 3. Pinning it here means the
+# high-signal configuration runs whatever the shipped default becomes, instead of
+# the suite quietly weakening the day that default flips. It must be process-wide
+# (the ENVIRONMENT property), not an in-process scope: the compile pool and its
+# threads are stood up at initialization, and their allocations are half the
+# point. DirectVulkan only - the memo this pins is DirectVulkan's.
+gtest_discover_tests(MobileGLIntegrationTest
+    TEST_PREFIX "DirectVulkan.AsyncCompile."
+    TEST_FILTER "XfbAfterClipDistanceScenario.*"
+    DISCOVERY_TIMEOUT 30
+    PROPERTIES
+        LABELS integration-gpu
+        TIMEOUT ${MGL_ITEST_TIMEOUT}
+        ENVIRONMENT "${MGL_ITEST_VULKAN_ASYNC_ENVIRONMENT}"
+)
+
+# A fourth registration, of the depth/stencil readback scenarios, with the ES
+# shader-sampling emulation forced on. Not paranoia - without it these scenarios are
+# UNFALSIFIABLE on the machines this suite runs on: OpenGL ES has no depth or stencil
+# readback in core, but Mesa accepts the reads anyway, so on llvmpipe every one of them
+# goes green through a native path that the Adreno device does not have. Deleting the
+# entire emulation left all of them passing. With the flag the native spellings are off
+# the table and only the path the device actually takes remains. DirectGLES only - the
+# emulation is DirectGLES's.
+gtest_discover_tests(MobileGLIntegrationTest
+    TEST_PREFIX "DirectGLES.ForcedDepthStencilEmulation."
+    TEST_FILTER "DepthStencilReadback*Scenario.*"
+    DISCOVERY_TIMEOUT 30
+    PROPERTIES
+        LABELS integration-gpu
+        TIMEOUT ${MGL_ITEST_TIMEOUT}
+        ENVIRONMENT "${MGL_ITEST_GLES_FORCED_DS_ENVIRONMENT}"
+)

MobileGL/MG_IntegrationTest/Harness/HeadlessGL.cpp

@@ -0,0 +1,761 @@
+// MobileGL - MobileGL/MG_IntegrationTest/Harness/HeadlessGL.cpp
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#include "HeadlessGL.h"
+
+#include <algorithm>
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <ostream>
+#include <sstream>
+
+#if defined(_WIN32)
+#define WIN32_LEAN_AND_MEAN
+#include <windows.h>
+#elif defined(__ANDROID__)
+#include <android/hardware_buffer.h>
+#include <android/native_window.h>
+#include <media/NdkImage.h>
+#include <media/NdkImageReader.h>
+#endif
+
+// MobileGL's own headers, in the order MobileGL/Includes.h uses them: GL/gl.h
+// first, then glcorearb.h for the 3.x+ entry points. This binary links
+// MobileGL_s, so every gl*/egl* below binds to MobileGL's implementation, not
+// to a system loader.
+#ifdef GLAPI
+#undef GLAPI
+#endif
+#include <EGL/egl.h>
+#define GL_GLEXT_PROTOTYPES
+#include <GL/gl.h>
+#include <GL/glcorearb.h>
+#undef GL_GLEXT_PROTOTYPES
+
+// The pre-flight below runs the whole EGL bring-up in a forked child, which is
+// the only construction that is actually predictive here: MobileGL ABORTS
+// (MOBILEGL_ASSERT -> SIGTRAP) rather than returning an error on an unusable
+// platform, so nothing the parent can call in-process is allowed to be wrong.
+#if !defined(_WIN32) && !defined(__APPLE__) && !defined(__ANDROID__) && __has_include(<sys/wait.h>)
+#define MGITEST_HAVE_FORK_PREFLIGHT 1
+#include <csignal>
+#include <ctime>
+#include <sys/resource.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <unistd.h>
+#else
+#define MGITEST_HAVE_FORK_PREFLIGHT 0
+#endif
+
+namespace MGITest {
+
+    namespace {
+        // Small enough that a readback is cheap, big enough that "top third" and
+        // "bottom third" are unambiguous. Non-square on purpose: a transposing
+        // bug cannot hide behind a square.
+        constexpr int kSurfaceWidth = 128;
+        constexpr int kSurfaceHeight = 96;
+
+#if defined(_WIN32)
+        HWND g_testWindow = nullptr;
+
+        HWND CreateTestWindow() {
+            static const wchar_t* const kClassName = L"MobileGLIntegrationTestWindow";
+            static bool registered = false;
+            if (!registered) {
+                WNDCLASSW windowClass{};
+                windowClass.lpfnWndProc = DefWindowProcW;
+                windowClass.hInstance = GetModuleHandleW(nullptr);
+                windowClass.lpszClassName = kClassName;
+                if (RegisterClassW(&windowClass) == 0 && GetLastError() != ERROR_CLASS_ALREADY_EXISTS) {
+                    return nullptr;
+                }
+                registered = true;
+            }
+            return CreateWindowExW(0, kClassName, L"MobileGL Integration Test", WS_OVERLAPPEDWINDOW,
+                                   CW_USEDEFAULT, CW_USEDEFAULT, kSurfaceWidth, kSurfaceHeight, nullptr, nullptr,
+                                   GetModuleHandleW(nullptr), nullptr);
+        }
+#elif defined(__ANDROID__)
+        AImageReader* g_imageReader = nullptr;
+        ANativeWindow* g_imageReaderWindow = nullptr;
+
+        void DrainImageReader(void*, AImageReader* reader) {
+            AImage* image = nullptr;
+            if (AImageReader_acquireNextImage(reader, &image) == AMEDIA_OK && image != nullptr) {
+                AImage_delete(image);
+            }
+        }
+
+        bool CreateImageReaderWindow() {
+            if (g_imageReaderWindow != nullptr) return true;
+            constexpr int kMaxImages = 4;
+            const media_status_t status = AImageReader_newWithUsage(
+                kSurfaceWidth, kSurfaceHeight, AIMAGE_FORMAT_RGBA_8888,
+                AHARDWAREBUFFER_USAGE_GPU_SAMPLED_IMAGE | AHARDWAREBUFFER_USAGE_GPU_COLOR_OUTPUT,
+                kMaxImages, &g_imageReader);
+            if (status != AMEDIA_OK || g_imageReader == nullptr) return false;
+
+            AImageReader_ImageListener listener = {nullptr, DrainImageReader};
+            AImageReader_setImageListener(g_imageReader, &listener);
+            if (AImageReader_getWindow(g_imageReader, &g_imageReaderWindow) != AMEDIA_OK ||
+                g_imageReaderWindow == nullptr) {
+                AImageReader_setImageListener(g_imageReader, nullptr);
+                AImageReader_delete(g_imageReader);
+                g_imageReader = nullptr;
+                return false;
+            }
+            ANativeWindow_acquire(g_imageReaderWindow);
+            return true;
+        }
+
+        void DestroyImageReaderWindow() {
+            if (g_imageReaderWindow != nullptr) {
+                ANativeWindow_release(g_imageReaderWindow);
+                g_imageReaderWindow = nullptr;
+            }
+            if (g_imageReader != nullptr) {
+                AImageReader_setImageListener(g_imageReader, nullptr);
+                AImageReader_delete(g_imageReader);
+                g_imageReader = nullptr;
+            }
+        }
+#endif
+
+        bool UseWindowSurface() {
+#if defined(_WIN32)
+            const char* value = std::getenv("MOBILEGL_ITEST_WINDOW_SURFACE");
+            return value != nullptr && value[0] != '\0' && std::strcmp(value, "0") != 0;
+#elif defined(__ANDROID__)
+            return true;
+#else
+            return false;
+#endif
+        }
+
+        std::string EnvOr(const char* name, const char* fallback) {
+            const char* value = std::getenv(name);
+            return (value != nullptr && value[0] != '\0') ? std::string(value) : std::string(fallback);
+        }
+
+        // A skip reason is only useful if it says which call failed AND why, so
+        // every bring-up step reports the EGL error it left behind.
+        std::string WithEglError(const char* what) {
+            std::ostringstream out;
+            out << what << " (eglGetError=0x" << std::hex << eglGetError() << ")";
+            return out.str();
+        }
+
+        // The EGL objects one bring-up produces.
+        struct EglBringUp {
+            void* display = nullptr;
+            void* surface = nullptr;
+            void* context = nullptr;
+            std::string renderer;
+        };
+
+        // The harness is headless BY CONSTRUCTION, on every machine: it must never
+        // reach a window system, not even where one happens to be running. This is
+        // not a CI accommodation - it is what keeps a developer's run and a CI run
+        // the same run. The lane was wired up green on a workstation and immediately
+        // died on the runner precisely because the workstation had a DISPLAY (WSLg)
+        // and took Mesa's x11 platform, while the runner has none; that divergence
+        // is the bug, and pinning the platform here is the fix for it.
+        //
+        // Mesa selects its EGL platform from EGL_PLATFORM at loader time, so this
+        // has to run before the first EGL call in the process (see EnsureHeadless
+        // callers). surfaceless is the platform with no window-system dependency at
+        // all; the surface this file then creates is still a pbuffer, which every
+        // platform supports and which the amendment to this rule requires as the
+        // fallback shape on desktop. Android instead supplies an AImageReader
+        // ANativeWindow. DISPLAY/WAYLAND_DISPLAY are cleared as well so that a
+        // driver that consults them directly cannot reintroduce the dependency
+        // behind EGL's back.
+        void EnsureHeadlessPlatform() {
+#if defined(__linux__) && !defined(__ANDROID__)
+            static bool done = false;
+            if (done) {
+                return;
+            }
+            done = true;
+            // An explicit EGL_PLATFORM from the operator still wins: pinning a
+            // platform is exactly how someone reproduces a platform-specific bug.
+            if (std::getenv("EGL_PLATFORM") == nullptr) {
+                setenv("EGL_PLATFORM", "surfaceless", 1);
+            }
+            unsetenv("DISPLAY");
+            unsetenv("WAYLAND_DISPLAY");
+#endif
+        }
+
+        // THE bring-up, in one function so the pre-flight child and the parent run
+        // literally the same sequence - a pre-flight that tests something narrower
+        // than what the parent will do is exactly the kind of "predictive" check
+        // that is not.
+        //
+        // Returns 0 on success, or the 1-based index of the step that failed, and
+        // fills outReason either way.
+        int RunEglBringUp(EglBringUp& out, std::string& outReason) {
+            // Belt and braces: the pre-flight child and the parent both enter here,
+            // and neither may be the first to touch EGL without this having run.
+            EnsureHeadlessPlatform();
+            EGLDisplay display = eglGetDisplay(EGL_DEFAULT_DISPLAY);
+            if (display == EGL_NO_DISPLAY) {
+                outReason = WithEglError("eglGetDisplay(EGL_DEFAULT_DISPLAY) returned EGL_NO_DISPLAY");
+                return 1;
+            }
+            EGLint major = 0, minor = 0;
+            if (eglInitialize(display, &major, &minor) != EGL_TRUE) {
+                outReason = WithEglError("eglInitialize failed: no usable display/driver on this machine");
+                return 2;
+            }
+            if (eglBindAPI(EGL_OPENGL_API) != EGL_TRUE) {
+                outReason = WithEglError("eglBindAPI(EGL_OPENGL_API) failed");
+                return 3;
+            }
+
+            const bool useWindowSurface = UseWindowSurface();
+            const EGLint configAttribs[] = {EGL_SURFACE_TYPE,
+                                            useWindowSurface ? EGL_WINDOW_BIT : EGL_PBUFFER_BIT,
+                                            EGL_RED_SIZE,
+                                            8,
+                                            EGL_GREEN_SIZE,
+                                            8,
+                                            EGL_BLUE_SIZE,
+                                            8,
+                                            EGL_ALPHA_SIZE,
+                                            8,
+                                            EGL_DEPTH_SIZE,
+                                            24,
+                                            EGL_RENDERABLE_TYPE,
+                                            EGL_OPENGL_BIT,
+                                            EGL_NONE};
+            EGLConfig config = nullptr;
+            EGLint configCount = 0;
+            if (eglChooseConfig(display, configAttribs, &config, 1, &configCount) != EGL_TRUE || configCount < 1) {
+                outReason = WithEglError(useWindowSurface
+                                             ? "eglChooseConfig found no window-capable RGBA8/D24 config"
+                                             : "eglChooseConfig found no pbuffer-capable RGBA8/D24 config");
+                return 4;
+            }
+
+            const EGLint contextAttribs[] = {EGL_CONTEXT_MAJOR_VERSION, 3, EGL_CONTEXT_MINOR_VERSION, 3, EGL_NONE};
+            EGLContext context = eglCreateContext(display, config, EGL_NO_CONTEXT, contextAttribs);
+            if (context == EGL_NO_CONTEXT) {
+                context = eglCreateContext(display, config, EGL_NO_CONTEXT, nullptr);
+            }
+            if (context == EGL_NO_CONTEXT) {
+                outReason = WithEglError("eglCreateContext failed: no desktop-GL context available");
+                return 5;
+            }
+
+            EGLSurface surface = EGL_NO_SURFACE;
+            if (useWindowSurface) {
+#if defined(_WIN32)
+                if (g_testWindow == nullptr) g_testWindow = CreateTestWindow();
+                if (g_testWindow == nullptr) {
+                    outReason = "failed to create the Windows integration-test window";
+                    return 6;
+                }
+                surface = eglCreateWindowSurface(display, config, g_testWindow, nullptr);
+#elif defined(__ANDROID__)
+                if (!CreateImageReaderWindow()) {
+                    outReason = "failed to create the Android AImageReader integration-test window";
+                    return 6;
+                }
+                surface = eglCreateWindowSurface(display, config, g_imageReaderWindow, nullptr);
+#endif
+            } else {
+                const EGLint pbufferAttribs[] = {EGL_WIDTH, kSurfaceWidth, EGL_HEIGHT, kSurfaceHeight, EGL_NONE};
+                surface = eglCreatePbufferSurface(display, config, pbufferAttribs);
+            }
+            if (surface == EGL_NO_SURFACE) {
+#if defined(__ANDROID__)
+                DestroyImageReaderWindow();
+#endif
+                outReason = WithEglError(useWindowSurface ? "eglCreateWindowSurface failed"
+                                                         : "eglCreatePbufferSurface failed");
+                return 6;
+            }
+            // The step that brings the whole backend up (DirectVulkan creates its
+            // instance, device and surface in here) and therefore the step that
+            // aborts instead of returning an error on an unusable platform.
+            if (eglMakeCurrent(display, surface, surface, context) != EGL_TRUE) {
+                outReason = WithEglError("eglMakeCurrent failed");
+                return 7;
+            }
+
+            const GLubyte* renderer = glGetString(GL_RENDERER);
+            if (renderer == nullptr) {
+                outReason = "glGetString(GL_RENDERER) returned null after eglMakeCurrent";
+                return 8;
+            }
+
+            out.display = display;
+            out.surface = surface;
+            out.context = context;
+            out.renderer = reinterpret_cast<const char*>(renderer);
+            outReason.clear();
+            return 0;
+        }
+
+        // Platform pre-flight, and the reason this module can claim to skip
+        // cleanly rather than merely hope to.
+        //
+        // MobileGL does not return errors when the platform is unusable - it
+        // ABORTS. MOBILEGL_ASSERT raises SIGTRAP, and the DirectVulkan bring-up
+        // asserts its way through instance, physical-device and surface creation
+        // inside eglMakeCurrent. So there is no in-process question the harness
+        // can ask that is guaranteed to be survivable, and the old form (dlopen
+        // the Vulkan loader, count physical devices, look for
+        // VK_EXT_headless_surface) was a guess at the abort conditions rather
+        // than a test of them: it named three of the ways bring-up can die and
+        // was silent about every other one, including every DirectGLES one.
+        //
+        // What is actually predictive is to run the bring-up itself somewhere a
+        // SIGTRAP is a datum instead of a crash. fork() gives exactly that: the
+        // child performs the identical sequence and _exit(0)s on success, and
+        // ANY non-zero exit or ANY signal in the parent's waitpid() means "this
+        // platform is unusable" - whatever the reason, including reasons nobody
+        // has thought of. Only then does the parent do the real bring-up.
+        //
+        // Returns an empty string when the platform survived a full bring-up.
+        std::string PreflightBringUp() {
+#if !MGITEST_HAVE_FORK_PREFLIGHT
+            // No fork(): let the in-process bring-up speak for itself, which is
+            // what this module did before. Windows/macOS are not CI targets for
+            // the headless scenarios.
+            return {};
+#else
+            int channel[2] = {-1, -1};
+            if (pipe(channel) != 0) {
+                return {}; // cannot pre-flight; fall through to the in-process attempt
+            }
+            // The child inherits our stdio buffers; flush so nothing is printed twice.
+            std::fflush(nullptr);
+            const pid_t child = fork();
+            if (child < 0) {
+                close(channel[0]);
+                close(channel[1]);
+                return {};
+            }
+            if (child == 0) {
+                close(channel[0]);
+                // No core suppression here, deliberately: when the child dies on a
+                // signal, the core IS the diagnosis (an rlimit that used to sit here
+                // made a CI-only crash undebuggable). Machines that do not want
+                // cores control that with the usual ulimit/core_pattern knobs.
+                std::fprintf(stderr, "[itest] pre-flight child: attempting a full EGL bring-up\n");
+                EglBringUp local;
+                std::string reason;
+                const int step = RunEglBringUp(local, reason);
+                if (!reason.empty()) {
+                    const std::size_t bytes = std::min<std::size_t>(reason.size(), 480);
+                    const ssize_t written = write(channel[1], reason.data(), bytes);
+                    (void)written;
+                }
+                close(channel[1]);
+                // _exit, never exit(): every atexit handler and static destructor
+                // in this address space belongs to the parent's copy of the world,
+                // and the child is holding a live context it must not tear down.
+                _exit(step);
+            }
+
+            close(channel[1]);
+            // Reap first, read after: the message is bounded well below the pipe
+            // buffer so the child can never block writing it, and polling the exit
+            // status is what lets a wedged child be killed instead of hanging the
+            // parent on a read that will never return.
+            constexpr int kPreflightTimeoutMs = 30000;
+            int status = 0;
+            int waitedMs = 0;
+            for (;;) {
+                const pid_t reaped = waitpid(child, &status, WNOHANG);
+                if (reaped == child) break;
+                if (reaped < 0) {
+                    close(channel[0]);
+                    return "waitpid on the EGL bring-up pre-flight child failed";
+                }
+                if (waitedMs >= kPreflightTimeoutMs) {
+                    kill(child, SIGKILL);
+                    (void)waitpid(child, &status, 0);
+                    close(channel[0]);
+                    std::ostringstream out;
+                    out << "the EGL bring-up wedged: a forked pre-flight child made no progress in "
+                        << kPreflightTimeoutMs / 1000 << "s and was killed";
+                    return out.str();
+                }
+                timespec nap{0, 10 * 1000 * 1000};
+                nanosleep(&nap, nullptr);
+                waitedMs += 10;
+            }
+
+            std::string childSays;
+            char buffer[512];
+            for (;;) {
+                const ssize_t got = read(channel[0], buffer, sizeof(buffer));
+                if (got <= 0) break;
+                childSays.append(buffer, static_cast<std::size_t>(got));
+            }
+            close(channel[0]);
+
+            if (WIFSIGNALED(status)) {
+                const int signalNumber = WTERMSIG(status);
+                const char* signalName = strsignal(signalNumber);
+                std::ostringstream out;
+                out << "the EGL bring-up ABORTS on this platform: a forked pre-flight child died on signal "
+                    << signalNumber << " (" << (signalName != nullptr ? signalName : "?") << ")";
+                if (!childSays.empty()) out << " after: " << childSays;
+                out << ". MobileGL asserts rather than returning an error here, so the scenarios would "
+                       "have taken the whole test binary down with them";
+                return out.str();
+            }
+            if (!WIFEXITED(status)) {
+                return "the EGL bring-up pre-flight child neither exited nor was signalled";
+            }
+            const int exitStatus = WEXITSTATUS(status);
+            if (exitStatus != 0) {
+                std::ostringstream out;
+                out << (childSays.empty() ? "the EGL bring-up failed" : childSays)
+                    << " (forked pre-flight child exit status " << exitStatus << ")";
+                return out.str();
+            }
+            return {};
+#endif
+        }
+    } // namespace
+
+    namespace {
+        bool EnvFlag(const char* name) {
+            const char* value = std::getenv(name);
+            return value != nullptr && value[0] != '\0' && std::strcmp(value, "0") != 0;
+        }
+    } // namespace
+
+    bool RequireGpu() {
+        return EnvFlag("MOBILEGL_ITEST_REQUIRE_GPU");
+    }
+
+    bool RequireHardwareGpu() {
+        return EnvFlag("MOBILEGL_ITEST_REQUIRE_HARDWARE_GPU");
+    }
+
+    std::ostream& operator<<(std::ostream& os, const Rgba8& c) {
+        os << "rgba(" << int(c.r) << "," << int(c.g) << "," << int(c.b) << "," << int(c.a) << ")";
+        return os;
+    }
+
+    Rgba8 Image::At(int x, int y) const {
+        if (x < 0 || y < 0 || x >= m_width || y >= m_height) {
+            return Rgba8{};
+        }
+        const std::size_t index = (static_cast<std::size_t>(y) * m_width + x) * 4;
+        return Rgba8{m_pixels[index], m_pixels[index + 1], m_pixels[index + 2], m_pixels[index + 3]};
+    }
+
+    const char* Image::ColorName(int x, int y) const {
+        const Rgba8 c = At(x, y);
+        const bool r = c.r > 160, g = c.g > 160, b = c.b > 160;
+        const bool nr = c.r < 96, ng = c.g < 96, nb = c.b < 96;
+        if (nr && ng && nb) return "black";
+        if (r && g && b) return "white";
+        if (r && ng && nb) return "red";
+        if (nr && g && nb) return "green";
+        if (nr && ng && b) return "blue";
+        if (r && g && nb) return "yellow";
+        return "other";
+    }
+
+    std::size_t Image::ByteDiffCount(const Image& other) const {
+        if (m_width != other.m_width || m_height != other.m_height) {
+            return std::max(m_pixels.size(), other.m_pixels.size());
+        }
+        std::size_t differing = 0;
+        for (std::size_t i = 0; i < m_pixels.size(); ++i) {
+            if (m_pixels[i] != other.m_pixels[i]) ++differing;
+        }
+        return differing;
+    }
+
+    std::string Image::QuadrantSignature() const {
+        if (m_width < 2 || m_height < 2) return "<empty>";
+        // Quadrant CENTRES, so a one-pixel rounding difference at a quadrant edge
+        // never decides the answer. Order is fixed and load-bearing: bottom-left,
+        // bottom-right, top-left, top-right.
+        const int leftX = m_width / 4;
+        const int rightX = m_width * 3 / 4;
+        const int bottomY = m_height / 4;
+        const int topY = m_height * 3 / 4;
+        std::ostringstream out;
+        out << ColorName(leftX, bottomY) << "," << ColorName(rightX, bottomY) << "," << ColorName(leftX, topY) << ","
+            << ColorName(rightX, topY);
+        return out.str();
+    }
+
+    RegionScan ScanRegion(const Image& image, int x0, int x1, int y0, int y1, const char* expectedColor) {
+        RegionScan scan;
+        x0 = std::max(x0, 0);
+        y0 = std::max(y0, 0);
+        x1 = std::min(x1, image.Width() - 1);
+        y1 = std::min(y1, image.Height() - 1);
+        for (int y = y0; y <= y1; ++y) {
+            for (int x = x0; x <= x1; ++x) {
+                ++scan.total;
+                const char* name = image.ColorName(x, y);
+                if (std::strcmp(name, expectedColor) == 0) continue;
+                ++scan.offenders;
+                if (scan.firstX < 0) {
+                    scan.firstX = x;
+                    scan.firstY = y;
+                    scan.firstColor = image.At(x, y);
+                    scan.firstColorName = name;
+                }
+            }
+        }
+        return scan;
+    }
+
+    ::testing::AssertionResult RegionIsMostly(const Image& image, int x0, int x1, int y0, int y1,
+                                              const char* expectedColor, double tolerance,
+                                              const std::string& when) {
+        const RegionScan scan = ScanRegion(image, x0, x1, y0, y1, expectedColor);
+        if (scan.total == 0) {
+            return ::testing::AssertionFailure()
+                   << when << ": region x[" << x0 << "," << x1 << "] y[" << y0 << "," << y1
+                   << "] is empty against a " << image.Width() << "x" << image.Height() << " readback";
+        }
+        const double offendingFraction = static_cast<double>(scan.offenders) / scan.total;
+        if (offendingFraction <= tolerance) {
+            return ::testing::AssertionSuccess();
+        }
+        return ::testing::AssertionFailure()
+               << when << ": region x[" << x0 << "," << x1 << "] y[" << y0 << "," << y1 << "] should be all "
+               << expectedColor << ", but " << scan.offenders << " of " << scan.total << " pixels ("
+               << static_cast<int>(offendingFraction * 100.0 + 0.5) << "%) are not; first offender at (" << scan.firstX
+               << "," << scan.firstY << ") is " << scan.firstColorName << " " << scan.firstColor;
+    }
+
+    HeadlessGL& HeadlessGL::Get() {
+        static HeadlessGL instance;
+        return instance;
+    }
+
+    HeadlessGL::HeadlessGL() {
+        // Before anything else in this process can reach EGL, and in particular
+        // before the pre-flight forks - the child must measure the same platform
+        // the parent will use.
+        EnsureHeadlessPlatform();
+        m_backendName = EnvOr("MOBILEGL_BACKEND_TYPE", "<unset>");
+        m_usable = BringUp();
+    }
+
+    bool HeadlessGL::BringUp() {
+        // Ask a disposable copy of this process first. Only if it survived does
+        // the real one try - see PreflightBringUp for why nothing weaker is
+        // predictive against a stack that aborts instead of returning errors.
+        const std::string preflightProblem = PreflightBringUp();
+        if (!preflightProblem.empty()) {
+            m_skipReason = preflightProblem;
+            return false;
+        }
+
+        // Same shape as DriverBench's boot_egl(), minus the dlopen: the provider
+        // is this binary. A pbuffer needs no window system, but MobileGL's own
+        // loader still has to reach a real driver underneath - and the child
+        // above just proved it can.
+        EglBringUp brought;
+        std::string reason;
+        if (RunEglBringUp(brought, reason) != 0) {
+            // The pre-flight passed and the parent's identical attempt did not.
+            // That is a real result, not a machine without a GPU, so say so: it
+            // means something is different between the two attempts (a leaked
+            // exclusive device, an environment the child did not have).
+            m_skipReason = reason + " - although an identical bring-up in a forked pre-flight child succeeded";
+            return false;
+        }
+
+        m_display = brought.display;
+        m_surface = brought.surface;
+        m_context = brought.context;
+        m_width = kSurfaceWidth;
+        m_height = kSurfaceHeight;
+        m_renderer = std::move(brought.renderer);
+        return true;
+    }
+
+    void HeadlessGL::EndFrame() {
+        if (!m_usable) return;
+        eglSwapBuffers(static_cast<EGLDisplay>(m_display), static_cast<EGLSurface>(m_surface));
+        ++m_frameIndex;
+    }
+
+    void HeadlessGL::ShutDown() {
+        if (!m_usable) return;
+        EGLDisplay display = static_cast<EGLDisplay>(m_display);
+        eglMakeCurrent(display, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
+        if (m_context != nullptr) eglDestroyContext(display, static_cast<EGLContext>(m_context));
+        if (m_surface != nullptr) eglDestroySurface(display, static_cast<EGLSurface>(m_surface));
+        eglTerminate(display);
+#if defined(_WIN32)
+        if (g_testWindow != nullptr) {
+            DestroyWindow(g_testWindow);
+            g_testWindow = nullptr;
+        }
+#elif defined(__ANDROID__)
+        DestroyImageReaderWindow();
+#endif
+        m_context = nullptr;
+        m_surface = nullptr;
+        m_display = nullptr;
+        m_usable = false;
+        m_skipReason = "the headless context has already been torn down";
+    }
+
+    // ---- scenario vocabulary ------------------------------------------------
+
+    namespace {
+        unsigned int CompileStage(GLenum stage, const char* source, std::string* outError) {
+            const GLuint shader = glCreateShader(stage);
+            glShaderSource(shader, 1, &source, nullptr);
+            glCompileShader(shader);
+            GLint compiled = 0;
+            glGetShaderiv(shader, GL_COMPILE_STATUS, &compiled);
+            if (compiled == GL_FALSE) {
+                char log[2048] = {};
+                GLsizei length = 0;
+                glGetShaderInfoLog(shader, sizeof(log) - 1, &length, log);
+                if (outError != nullptr) {
+                    *outError = std::string(stage == GL_VERTEX_SHADER ? "vertex" : "fragment") +
+                                " shader failed to compile: " + log;
+                }
+                glDeleteShader(shader);
+                return 0;
+            }
+            return shader;
+        }
+    } // namespace
+
+    unsigned int CompileProgram(const char* vertexSource, const char* fragmentSource, std::string* outError) {
+        const GLuint vs = CompileStage(GL_VERTEX_SHADER, vertexSource, outError);
+        if (vs == 0) return 0;
+        const GLuint fs = CompileStage(GL_FRAGMENT_SHADER, fragmentSource, outError);
+        if (fs == 0) {
+            glDeleteShader(vs);
+            return 0;
+        }
+        const GLuint program = glCreateProgram();
+        glAttachShader(program, vs);
+        glAttachShader(program, fs);
+        // Pinned rather than queried so the scenarios can set up a VAO without a
+        // round trip, and so a driver that reorders attributes cannot change what
+        // the test means.
+        glBindAttribLocation(program, 0, "aPos");
+        glBindAttribLocation(program, 1, "aColor");
+        glLinkProgram(program);
+        glDeleteShader(vs);
+        glDeleteShader(fs);
+        GLint linked = 0;
+        glGetProgramiv(program, GL_LINK_STATUS, &linked);
+        if (linked == GL_FALSE) {
+            char log[2048] = {};
+            GLsizei length = 0;
+            glGetProgramInfoLog(program, sizeof(log) - 1, &length, log);
+            if (outError != nullptr) *outError = std::string("program failed to link: ") + log;
+            glDeleteProgram(program);
+            return 0;
+        }
+        return program;
+    }
+
+    ColorFbo MakeColorFbo(int width, int height) {
+        ColorFbo target;
+        target.width = width;
+        target.height = height;
+        glGenTextures(1, &target.texture);
+        glBindTexture(GL_TEXTURE_2D, target.texture);
+        glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA8, width, height, 0, GL_RGBA, GL_UNSIGNED_BYTE, nullptr);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+        glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+        glBindTexture(GL_TEXTURE_2D, 0);
+
+        glGenFramebuffers(1, &target.fbo);
+        glBindFramebuffer(GL_FRAMEBUFFER, target.fbo);
+        glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, target.texture, 0);
+        const GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER);
+        glBindFramebuffer(GL_FRAMEBUFFER, 0);
+        if (status != GL_FRAMEBUFFER_COMPLETE) {
+            DestroyColorFbo(target);
+        }
+        return target;
+    }
+
+    void DestroyColorFbo(ColorFbo& target) {
+        if (target.fbo != 0) glDeleteFramebuffers(1, &target.fbo);
+        if (target.texture != 0) glDeleteTextures(1, &target.texture);
+        target.fbo = 0;
+        target.texture = 0;
+    }
+
+    void BindDefaultFramebuffer() {
+        glBindFramebuffer(GL_FRAMEBUFFER, 0);
+        glViewport(0, 0, HeadlessGL::Get().Width(), HeadlessGL::Get().Height());
+    }
+
+    void BindFbo(const ColorFbo& target) {
+        glBindFramebuffer(GL_FRAMEBUFFER, target.fbo);
+        glViewport(0, 0, target.width, target.height);
+    }
+
+    void ClearTo(float r, float g, float b, float a) {
+        glClearColor(r, g, b, a);
+        glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+    }
+
+    Image ReadPixels(int width, int height) {
+        return ReadPixelsRect(0, 0, width, height);
+    }
+
+    Image ReadPixelsRect(int x, int y, int width, int height) {
+        Image image(width, height);
+        glPixelStorei(GL_PACK_ALIGNMENT, 1);
+        glReadPixels(x, y, width, height, GL_RGBA, GL_UNSIGNED_BYTE, image.Data());
+        return image;
+    }
+
+    unsigned int FirstGLError() {
+        const GLenum first = glGetError();
+        if (first == GL_NO_ERROR) return GL_NO_ERROR;
+        // Drain, bounded: a broken stack must not turn an error check into a hang.
+        for (int i = 0; i < 64 && glGetError() != GL_NO_ERROR; ++i) {}
+        return first;
+    }
+
+    const char* GLErrorName(unsigned int error) {
+        switch (error) {
+        case GL_NO_ERROR:
+            return "GL_NO_ERROR";
+        case GL_INVALID_ENUM:
+            return "GL_INVALID_ENUM";
+        case GL_INVALID_VALUE:
+            return "GL_INVALID_VALUE";
+        case GL_INVALID_OPERATION:
+            return "GL_INVALID_OPERATION";
+        case GL_OUT_OF_MEMORY:
+            return "GL_OUT_OF_MEMORY";
+        case GL_INVALID_FRAMEBUFFER_OPERATION:
+            return "GL_INVALID_FRAMEBUFFER_OPERATION";
+        default:
+            return "GL_<unknown>";
+        }
+    }
+
+} // namespace MGITest