[Test] (SelfTest, POST): probe upload/draw ordering in persistently mapped vertex arenas - native GLES controls distinguish mapped destination corruption from staging and synchronization failures

Add the persistent buffer ordering probe to the GLES POST's known-driver-bug inventory. Queue updates and draws into independent FBOs before reading them back, covering SubData and copies from both coherent persistent and ordinary staging buffers. Retry fresh mapped allocations for intermittent corruption.

Require passing never-mapped and fully serialized controls before reporting a finding. Include finish-before, map-then-unmap and barrier diagnostics, preserve caller GL state, and treat setup, allocation and GL errors as inconclusive. This adds detection and reporting only; no rendering workaround is enabled.

Validation: all 46 DriverBugProbesTest tests pass, including 11 new ordering, control, collector and cleanup cases. The Android API 26 / NDK 27 native probe detects all three upload paths on Mali-G1-Ultra r54p1 with clean controls and no GL errors. llvmpipe reports no finding after 240 mapped FBO checks per upload path.

.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
@@ -9,7 +13,23 @@ fi
 case_name="$1"
 fixture_dir="${2:-tools/trace_replay/fixtures}"
 python_bin="${PYTHON:-python3}"
-mirror_base="${MOBILEGL_TRACE_FIXTURE_MIRROR_BASE:-https://repo.miawa.cn/mgl/tools/trace_replay/fixtures}"
+# Fixture mirrors, tried in order before falling back to Git LFS. Override the
+# whole list with MOBILEGL_TRACE_FIXTURE_MIRROR_BASES (whitespace separated);
+# MOBILEGL_TRACE_FIXTURE_MIRROR_BASE still works and is tried first.
+default_mirror_bases=(
+  "https://git.hit.moe/swung0x48/MobileGL/media/branch/dev/tools/trace_replay/fixtures"
+  "https://repo.miawa.cn/mgl/tools/trace_replay/fixtures"
+)
+if [ -n "${MOBILEGL_TRACE_FIXTURE_MIRROR_BASES:-}" ]; then
+  read -r -a mirror_bases <<< "${MOBILEGL_TRACE_FIXTURE_MIRROR_BASES}"
+else
+  mirror_bases=("${default_mirror_bases[@]}")
+fi
+if [ -n "${MOBILEGL_TRACE_FIXTURE_MIRROR_BASE:-}" ]; then
+  mirror_bases=("${MOBILEGL_TRACE_FIXTURE_MIRROR_BASE}" "${mirror_bases[@]}")
+fi
+# Optional bearer token for mirrors that require authentication (private Gitea).
+mirror_token="${MOBILEGL_TRACE_FIXTURE_MIRROR_TOKEN:-}"
 download_attempts="${MOBILEGL_TRACE_FIXTURE_DOWNLOAD_ATTEMPTS:-5}"
 retry_delay="${MOBILEGL_TRACE_FIXTURE_RETRY_DELAY:-2}"
 
@@ -30,7 +50,8 @@ fixture_list="$("${python_bin}" tools/trace_replay/trace_cases.py \
   --format fixture-files \
   --case "${case_name}" \
   --fixture-root "${fixture_dir}")"
-mapfile -t files <<< "${fixture_list}"
+# Strip CR so the script also works when python emits CRLF (Git Bash on Windows).
+mapfile -t files < <(printf '%s\n' "${fixture_list}" | tr -d '\r')
 
 include="$(IFS=,; echo "${files[*]}")"
 if [ "${case_name}" = "OpenRA" ]; then
@@ -45,57 +66,6 @@ if [ "${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"
@@ -106,6 +76,7 @@ fetch_file_from_mirror() {
   local attempt
   local partial_size
   local curl_status
+  local curl_auth
 
   metadata="$(get_lfs_metadata "${file}")" || return 1
   read -r expected_oid expected_size <<< "${metadata}"
@@ -136,7 +107,11 @@ fetch_file_from_mirror() {
       echo "Starting mirror download for ${file} (attempt ${attempt}/${download_attempts})"
     fi
 
-    if curl -L --fail --show-error --continue-at - --output "${tmp_file}" "${url}"; then
+    curl_auth=()
+    if [ -n "${mirror_token}" ]; then
+      curl_auth=(--header "Authorization: token ${mirror_token}")
+    fi
+    if curl -L --fail --show-error --continue-at - "${curl_auth[@]}" --output "${tmp_file}" "${url}"; then
       if verify_fixture_file "${tmp_file}" "${file}" "${expected_oid}" "${expected_size}"; then
         mv "${tmp_file}" "${file}"
         return 0
@@ -179,26 +154,42 @@ fetch_file_from_mirror() {
   return 1
 }
 
+# Files no mirror could serve, even after retrying every mirror. Only these fall
+# back to Git LFS, so a mirror that served the rest of the case still spares
+# GitHub the bandwidth for those files.
+mirror_failures=()
+
 fetch_from_mirror() {
   mkdir -p "${fixture_dir}"
   for file in "${files[@]}"; do
     local name
     local url
+    local base
+    local fetched=0
     name="$(basename "${file}")"
-    url="${mirror_base%/}/${name}"
-    echo "Fetching trace fixture from mirror: ${url}"
-    if ! fetch_file_from_mirror "${file}" "${url}"; then
-      return 1
+    for base in "${mirror_bases[@]}"; do
+      url="${base%/}/${name}"
+      echo "Fetching trace fixture from mirror: ${url}"
+      if fetch_file_from_mirror "${file}" "${url}"; then
+        fetched=1
+        break
+      fi
+      echo "Mirror did not serve ${name}; trying the next mirror" >&2
+    done
+    if [ "${fetched}" -ne 1 ]; then
+      mirror_failures+=("${file}")
     fi
   done
+  [ "${#mirror_failures[@]}" -eq 0 ]
 }
 
 if fetch_from_mirror; then
   echo "Fetched trace fixture files for ${case_name} from mirror: ${include}"
 else
-  echo "Mirror fetch failed for ${case_name}; falling back to Git LFS: ${include}"
+  fallback_include="$(IFS=,; echo "${mirror_failures[*]}")"
+  echo "All mirrors failed for ${#mirror_failures[@]} of ${#files[@]} file(s) of ${case_name}; falling back to Git LFS: ${fallback_include}"
   git lfs install --local
-  git lfs pull --include="${include}" --exclude=""
+  git lfs pull --include="${fallback_include}" --exclude=""
 fi
 
 for file in "${files[@]}"; do

.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/scripts/validate-plugin-apks.sh

@@ -44,12 +44,12 @@ require 'key:MOBILEGL_BACKEND_TYPE' "$plugin_resource_text" 'V2 backend variable
 require 'defaultValue:DirectGLES' "$plugin_resource_text" 'V2 DirectGLES default'
 require 'DirectVulkan' "$plugin_resource_text" 'V2 DirectVulkan option'
 require 'key:MOBILEGL_DISABLE_TIMERQUERY' "$plugin_resource_text" 'V2 timer-query toggle'
-require 'key:MOBILEGL_DISABLE_SUBGROUP' "$plugin_resource_text" 'V2 Vulkan subgroup toggle'
+require 'key:MOBILEGL_MAGMA_DISABLE_SUBGROUP' "$plugin_resource_text" 'V2 Vulkan subgroup toggle'
 require 'key:MOBILEGL_MAGMA_R11G11B10F_FALLBACK' "$plugin_resource_text" 'V2 Magma format fallback toggle'
 require 'key:MOBILEGL_MAGMA_FRAMESINFLIGHT' "$plugin_resource_text" 'V2 Magma frames-in-flight setting'
-require 'key:MOBILEGL_AVOID_SAMPLER_MIPMAP_MIN_FILTER' "$plugin_resource_text" 'V2 sampler workaround toggle'
+require 'key:MOBILEGL_ESPRYT_AVOID_SAMPLER_MIPMAP_MIN_FILTER' "$plugin_resource_text" 'V2 sampler workaround toggle'
 require 'key:MOBILEGL_COHERENT_AS_FLUSH' "$plugin_resource_text" 'V2 coherent-as-flush toggle'
-require 'key:MOBILEGL_USE_ANGLE' "$plugin_resource_text" 'V2 ANGLE toggle'
+require 'key:MOBILEGL_ESPRYT_USE_ANGLE' "$plugin_resource_text" 'V2 ANGLE toggle'
 
 if [[ $(grep -Fc 'fclPlugin_V2' <<<"$plugin_manifest") -ne 1 ]]; then
   echo '::error::Plugin manifest must expose exactly one V2 descriptor' >&2

.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:
-        backend:
-          - name: DirectGLES
-            gpu: software
-          - name: DirectVulkan
-            gpu: lavapipe
-        case: ${{ fromJSON(needs.trace-cases.outputs.android) }}
+      matrix: ${{ fromJSON(needs.trace-cases.outputs.android) }}
     steps:
       - name: Set Swap Space
         uses: pierotofy/set-swap-space@v1.0
@@ -367,9 +417,12 @@ jobs:
 
       - name: Retrace and validate
         env:
-          MOBILEGL_USE_ANGLE: ${{ matrix.backend.name == 'DirectGLES' && '1' || '0' }}
+          MOBILEGL_ESPRYT_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_MAGMA_FIX_ITERATIONRP_SUBGROUP_SCRATCH: ${{ matrix.backend.name == 'DirectVulkan' && matrix.case.name == 'minecraft-1.21.4-fabric-iris-iterationrp-in-world' && '1' || '0' }}
+          MOBILEGL_MAGMA_DERIVE_NUM_SUBGROUPS: ${{ matrix.backend.name == 'DirectVulkan' && matrix.case.name == 'minecraft-1.21.4-fabric-iris-iterationrp-in-world' && '1' || '0' }}
+          MOBILEGL_MAGMA_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()
@@ -531,22 +612,41 @@ 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
-                echo "Retaining ${artifact_name} (${artifact_id}) for failed retrace."
-                ((retained += 1))
-                continue
+                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
 
             echo "Deleting ${artifact_name} (${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 +186,115 @@ 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_MAGMA_FIX_ITERATIONRP_SUBGROUP_SCRATCH: "1"
+          MOBILEGL_MAGMA_DERIVE_NUM_SUBGROUPS: "1"
+          MOBILEGL_MAGMA_ITERATIONRP_FIX_BARRIER: "1"
+        run: |
+          ulimit -c unlimited
+          sudo sysctl -w kernel.core_pattern='/tmp/core.%e.%p'
+          # Second, filtered pass: with the range-invalidating map flush disabled,
+          # the buffer scenarios run on the upload ring's staged-copy tier - which
+          # the default pass never reaches (the map tier absorbs every flush on
+          # Mesa), so without this the Mali fallback tier would have zero CI
+          # coverage. The flag is NOT baked into the ctest ENVIRONMENT properties,
+          # so an inline env reaches the test processes (unlike the ICD pin above).
+          if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" = "true" ]; then
+            ctest -V -L integration-gpu --no-tests=error
+            MOBILEGL_ESPRYT_DISABLE_INVALIDATE_FLUSH=1 ctest -V -L integration-gpu \
+              -R 'Buffer|Readback|Atomic|Ssbo|Arena' --no-tests=error
+          else
+            ctest --output-on-failure -L integration-gpu --no-tests=error
+            MOBILEGL_ESPRYT_DISABLE_INVALIDATE_FLUSH=1 ctest --output-on-failure -L integration-gpu \
+              -R 'Buffer|Readback|Atomic|Ssbo|Arena' --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 +307,13 @@ 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 +338,18 @@ 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 +357,10 @@ jobs:
       - build-linux
       - test
       - benchmark
+      - integration
+    permissions:
+      actions: write
+      contents: read
     env:
       BUILD_DIR: build-retrace
       CCACHE_BASEDIR: ${{ github.workspace }}
@@ -236,12 +385,11 @@ 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 +455,21 @@ 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 +502,9 @@ jobs:
     needs:
       - test
       - benchmark
+      - integration
     outputs:
+      matrix: ${{ steps.trace-cases.outputs.matrix }}
       names: ${{ steps.trace-cases.outputs.names }}
     steps:
       - name: Checkout repo
@@ -347,7 +512,9 @@ 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 +529,41 @@ 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 +593,7 @@ jobs:
     strategy:
       fail-fast: false
       max-parallel: 4
-      matrix:
-        backend:
-          - DirectGLES
-          - DirectVulkan
-        case: ${{ fromJSON(needs.trace-cases.outputs.names) }}
+      matrix: ${{ fromJSON(needs.trace-cases.outputs.matrix) }}
 
     steps:
       - name: Set Swap Space
@@ -452,11 +647,36 @@ 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' ] \
+             && [ '${{ matrix.case }}' = 'minecraft-1.21.4-fabric-iris-iterationrp-in-world' ]; then
+            export MOBILEGL_MAGMA_FIX_ITERATIONRP_SUBGROUP_SCRATCH=1
+            export MOBILEGL_MAGMA_DERIVE_NUM_SUBGROUPS=1
+            export MOBILEGL_MAGMA_ITERATIONRP_FIX_BARRIER=1
+          fi
+          # The blended depth-write quirk auto-enables only on Qualcomm, which no CI
+          # runner has, so force it on for the OIT case it exists to fix. ForceOn
+          # bypasses only the vendor gate, so this exercises the real strip on
+          # lavapipe. The Android AVD lane deliberately leaves it off, keeping the
+          # unstripped path covered for the same trace.
+          if [ '${{ matrix.backend }}' = 'DirectVulkan' ] \
+             && [ '${{ matrix.case }}' = 'improved-transparency-minecraft-26.3' ]; then
+            export MOBILEGL_MAGMA_DISABLE_BLENDED_DEPTH_WRITE=1
+          fi
           ctest -V --no-tests=error -R '^MobileGLTraceReplay\.${{ 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

@@ -25,3 +25,6 @@ MobileGL/MG*/cmake-build*
 /android-plugin/app/src/trace/jniLibs
 /android-plugin/local.properties
 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
 
@@ -180,27 +266,63 @@ 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
+    MobileGL/MG_Util/ShaderTranspiler/TranslationCache.cpp
     MobileGL/MG_Util/ShaderTranspiler/glslang/TMglGlslIoResolver.cpp
     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/FlattenFloat64StorageBlockPass.cpp
     MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/LowerDrawParametersPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/LowerViewportIndexPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/PackDoubleVertexInputsPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/FlattenXfbInterfaceBlocksPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/UniquifyIoBlockNamesPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/StripIoBlockLocationsPass.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/Lower1DSampledImagesPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/BakeImageFormatsPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/WidenImageFormatsPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/ClampMultisampleFetchPass.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/ShaderTranspiler/SpirvPasses/LegalizeResourceArrayIndexPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/FlattenAtomicCounterBlockPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/DemotePointSizePass.cpp
 
     MobileGL/MG_Util/BackendLoaders/OpenGL/Loader.cpp
     MobileGL/MG_Util/BackendLoaders/Vulkan/Loader.cpp
 
+    MobileGL/MG_Util/SelfTest/DriverBugProbes.cpp
+    MobileGL/MG_Util/SelfTest/PersistentBufferOrderingProbe.cpp
     MobileGL/MG_Util/SelfTest/DriverPost.cpp
+    MobileGL/MG_Util/SelfTest/DriverPostIterationRPWitness.cpp
+    MobileGL/MG_Util/SelfTest/PrimitivesGeneratedNoXfbProbe.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
@@ -214,7 +336,10 @@ 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/Debug/GL_Debug.cpp
     MobileGL/MG_Impl/GLImpl/Texture/Validators.cpp
     MobileGL/MG_Impl/GLImpl/Texture/ProxyTexture.cpp
     MobileGL/MG_Impl/GLImpl/VertexArray/GL_VertexArray.cpp
@@ -236,6 +361,7 @@ 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
@@ -271,10 +397,17 @@ set(SOURCE_FILES
     MobileGL/MG_State/GLState/TextureState/TextureObject2DCube.cpp
     MobileGL/MG_State/GLState/TextureState/TextureObject3D.cpp
     MobileGL/MG_State/GLState/TextureState/TextureObjectBuffer.cpp
+    MobileGL/MG_State/GLState/TextureState/TextureObjectView.cpp
     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/ProgramTranslationCache.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
@@ -297,9 +430,22 @@ endif()
 if (ANDROID)
     list(APPEND SOURCE_FILES
         MobileGL/MG_Util/SelfTest/DriverPostJni.cpp
+        MobileGL/MG_Util/SelfTest/DriverBenchJni.cpp
     )
 endif()
 
+if (WIN32)
+    list(APPEND SOURCE_FILES
+        MobileGL/MG_Impl/WGLImpl/WGLImpl.cpp
+        MobileGL/MG_Impl/WGLImpl/Exporting/Definitions.cpp
+    )
+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
@@ -309,12 +455,17 @@ 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}")
@@ -326,12 +477,24 @@ 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(${CMAKE_PROJECT_NAME} SHARED 
+add_library(${CMAKE_PROJECT_NAME} SHARED
     ${SOURCE_FILES}
 )
 
+if (WIN32)
+    # The wgl* entry points are exported via .def (see the comment in wgl.def);
+    # only the shared library links it.
+    target_sources(${CMAKE_PROJECT_NAME} PRIVATE
+        MobileGL/MG_Impl/WGLImpl/Exporting/wgl.def
+    )
+endif()
+
 if (CMAKE_BUILD_TYPE STREQUAL "Debug")
     set_target_properties(${CMAKE_PROJECT_NAME} PROPERTIES
         C_VISIBILITY_PRESET default
@@ -374,6 +537,18 @@ if(UNIX AND NOT APPLE AND NOT ANDROID)
     endforeach()
 endif()
 
+if(WIN32)
+    # Drop-in for the classic GL loader path: a copy named opengl32.dll placed
+    # next to a host executable is what LoadLibrary("opengl32.dll") and gdi32's
+    # pixel-format forwarding will resolve.
+    add_custom_command(TARGET ${CMAKE_PROJECT_NAME} POST_BUILD
+        COMMAND ${CMAKE_COMMAND} -E copy_if_different
+                "$<TARGET_FILE:${CMAKE_PROJECT_NAME}>"
+                "$<TARGET_FILE_DIR:${CMAKE_PROJECT_NAME}>/opengl32.dll"
+        COMMENT "Creating opengl32.dll drop-in copy"
+    )
+endif()
+
 if(NOT ANDROID)
     add_library(${CMAKE_PROJECT_NAME}_s STATIC 
         ${SOURCE_FILES}
@@ -425,8 +600,21 @@ if (ANDROID)
 endif()
 
 if (APPLE AND NOT MOBILEGL_IOS)
+    # MobileGL statically embeds glslang, SPIRV-Tools, and SPIRV-Cross. When
+    # this dylib is injected with DYLD_INSERT_LIBRARIES, exporting those C++
+    # symbols interposes incompatible copies embedded by host libraries such
+    # as shaderc. Keep only the public GL/EGL/CGL loader surface globally
+    # visible; GetProcAddress can still return pointers to hidden internals.
+    set(MOBILEGL_MACOS_EXPORTED_SYMBOLS
+        "${CMAKE_CURRENT_SOURCE_DIR}/MobileGL/MG_Impl/DyldInterpose/ExportedSymbols.txt")
+    target_link_options(${CMAKE_PROJECT_NAME} PRIVATE
+        "LINKER:-exported_symbols_list,${MOBILEGL_MACOS_EXPORTED_SYMBOLS}")
+    set_property(TARGET ${CMAKE_PROJECT_NAME} APPEND PROPERTY
+        LINK_DEPENDS "${MOBILEGL_MACOS_EXPORTED_SYMBOLS}")
+
     target_link_libraries(${CMAKE_PROJECT_NAME} PUBLIC
         "-framework Cocoa"
+        "-framework CoreVideo"
         "-framework QuartzCore"
         "-framework Foundation"
         "-framework OpenGL"
@@ -434,6 +622,7 @@ if (APPLE AND NOT MOBILEGL_IOS)
     if(TARGET ${CMAKE_PROJECT_NAME}_s)
         target_link_libraries(${CMAKE_PROJECT_NAME}_s PUBLIC
             "-framework Cocoa"
+            "-framework CoreVideo"
             "-framework QuartzCore"
             "-framework Foundation"
             "-framework OpenGL"
@@ -489,6 +678,12 @@ 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()
@@ -497,3 +692,10 @@ 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

@@ -14,12 +14,48 @@ namespace MobileGL::MG_Config {
     inline const String ProjectName = "MobileGL";
     inline const String CoreName = "MobileGL Core";
     inline const String CoreVendor = "MobileGL-Dev (BZLZHH, Swung0x48, Tungsten)";
-    inline const Version CoreVersion = {26, 7, 0, "-dev", VersionType::Development};
+    inline const Version CoreVersion = {26, 8, 0, "-dev", VersionType::Development};
     inline const VersionStringFormatAttrib DefaultVersionStringFormatAttrib = {2, 2, 0, true, true};
     inline const Uint64 CacheVersion = 0;
 
     extern BackendType ActiveBackendType;
 
+    // Tri-state override for device-specific quirks: Auto lets the detected device decide,
+    // ForceOn/ForceOff bypass the detection in either direction. ForceOn only bypasses the
+    // device gate - each quirk keeps its structural safety checks.
+    enum class QuirkOverride : Uint8 {
+        Auto = 0,
+        ForceOn,
+        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"
@@ -33,21 +69,102 @@ namespace MobileGL::MG_Config {
     struct FeaturesTable {
         // MOBILEGL_DISABLE_TIMERQUERY: do not advertise or use GPU timer queries.
         Bool DisableTimerQuery = false;
-        // MOBILEGL_USE_ANGLE: load ANGLE EGL/GLES libraries.
-        Bool UseAngle = false;
+        // MOBILEGL_ESPRYT_ENABLE_TEXTURE_VIEW: advertise GL_ARB_texture_view on DirectGLES when
+        // the host ES driver has EXT/OES_texture_view. Off by default: the host extension is
+        // present on Adreno 830 and the functional half of KHR-GL4{2,3}.texture_view still fails
+        // there, because the view's ES internalformat is normalized independently of the storage
+        // it aliases (see BackendObject_DirectGLES::BuildAdvertisedExtensions). The flag exists
+        // so that work can be done without editing the gate.
+        Bool EsprytEnableTextureView = 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_ESPRYT_USE_ANGLE: load ANGLE EGL/GLES libraries.
+        Bool EsprytUseAngle = 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.
-        Bool DisableSubgroup = false;
+        // MOBILEGL_MAGMA_DISABLE_SUBGROUP: force-disable Vulkan shader subgroup support,
+        // including the opt-in emulated compute path below.
+        Bool MagmaDisableSubgroup = 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
+        // MagmaFixIterationRPSubgroupScratch below instead). Off by default.
+        Bool MagmaEmulateSubgroup = false;
+        // MOBILEGL_MAGMA_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 MagmaFixIterationRPSubgroupScratch = QuirkOverride::Auto;
+        // MOBILEGL_MAGMA_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 MagmaIterationRPFixBarrier = false;
+        // MOBILEGL_MAGMA_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 MagmaDeriveNumSubgroups = 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.
         Uint32 MagmaFramesInFlight = 3;
-        // MOBILEGL_AVOID_SAMPLER_MIPMAP_MIN_FILTER: avoid mipmap min filters in samplers,
+        // MOBILEGL_ESPRYT_AVOID_SAMPLER_MIPMAP_MIN_FILTER: avoid mipmap min filters in samplers,
         // resolves certain rendering bugs on ANGLE + llvmpipe.
-        Bool AvoidSamplerMipmapMinFilter = false;
+        Bool EsprytAvoidSamplerMipmapMinFilter = false;
+        // MOBILEGL_ESPRYT_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 EsprytAvoidExplicitLodBias = false;
+        // MOBILEGL_ESPRYT_UNLOCATED_IO_BLOCKS: emit a tessellation/geometry program's
+        // inter-stage interface blocks WITHOUT their layout(location=) qualifier, letting ES
+        // match them by block name and member sequence instead. The Mali ES driver delivers
+        // nothing at all through a located block once a tessellation or geometry stage is in
+        // the pipeline; the driver POST measures that and turns this on by itself, so Auto is
+        // the right setting everywhere. ForceOn exists so the emulation can be exercised on a
+        // healthy driver - which is what the integration lane does, since llvmpipe and
+        // lavapipe carry a located block correctly and would otherwise never run this code -
+        // and ForceOff is the negative control. See StripIoBlockLocationsPass.
+        QuirkOverride EsprytUnlocatedIoBlocks = QuirkOverride::Auto;
+        // MOBILEGL_POINT_SIZE_DEMOTION: demote gl_PointSize out of tessellation/geometry
+        // stages into an ordinary varying (ShaderCompiler::
+        // DemoteTessellationGeometryPointSizeForProgram) instead of declining such programs
+        // on a device that advertises neither EXT/OES_tessellation_point_size /
+        // geometry_point_size (DirectGLES) nor shaderTessellationAndGeometryPointSize
+        // (DirectVulkan). Auto arms it exactly where the detection says the capability is
+        // absent, which is the right setting everywhere. ForceOn exists so the demotion can
+        // be exercised on a healthy driver - llvmpipe and lavapipe host the built-in
+        // natively and would otherwise never run this code, which is what the pinned
+        // integration lane uses - and ForceOff restores the plain declines (escape hatch /
+        // negative control). Cross-backend by design: the demotion runs in the shared
+        // phase-B chain, so one switch covers both. See DemotePointSizePass.
+        QuirkOverride PointSizeDemotion = QuirkOverride::Auto;
         // 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
@@ -57,16 +174,148 @@ namespace MobileGL::MG_Config {
         Bool CoherentAsFlush = false;
         // MOBILEGL_TRACE_SKIP_AUTODESTROY: skip teardown in the ELF destructor (Init.cpp).
         Bool TraceSkipAutodestroy = false;
-        // MOBILEGL_DISABLE_UBO_RING: force the DirectGLES global-UBO upload back to the
+        // MOBILEGL_ESPRYT_DISABLE_UBO_RING: force the DirectGLES global-UBO upload back to the
         // per-draw glBufferSubData path instead of the persistent-mapped ring allocator
         // (negative control / driver-bug escape hatch).
-        Bool DisableUboRing = false;
+        Bool EsprytDisableUboRing = false;
+        // MOBILEGL_ESPRYT_DISABLE_UNPACK_RING: force DirectGLES texture uploads back to
+        // glTexSubImage from the client pointer instead of staging them through the
+        // persistent-mapped unpack-PBO ring (negative control / driver-bug escape
+        // hatch).
+        Bool EsprytDisableUnpackRing = false;
+        // MOBILEGL_ESPRYT_DISABLE_UPLOAD_RING: force DirectGLES app buffer updates
+        // (glBufferSubData / map flushes) back to the immediate driver upload instead
+        // of queueing them for the staged-copy flush through the persistent-mapped
+        // upload ring (negative control / driver-bug escape hatch; the immediate
+        // upload stalls on drivers that resolve the WAR hazard on the CPU, e.g. Mali).
+        Bool EsprytDisableUploadRing = false;
+        // MOBILEGL_ESPRYT_DISABLE_INVALIDATE_FLUSH: skip the glMapBufferRange(WRITE |
+        // INVALIDATE_RANGE) tier of the DirectGLES pending-range flush and go straight
+        // to the upload ring's staged glCopyBufferSubData (negative control / escape
+        // hatch for a driver whose range-invalidating map misbehaves). The map tier is
+        // what keeps a partial write into a large in-flight buffer priced by the RANGE:
+        // on Mali both the immediate glBufferSubData and a staged copy into a busy
+        // mutable store ghost the whole destination on the CPU.
+        Bool EsprytDisableInvalidateFlush = false;
+        // MOBILEGL_DISABLE_LARGE_BUFFER_ADOPTION: keep mesh-arena-sized buffer stores
+        // (>= 16MiB) on the CPU-shadow model instead of backing them with the backend's
+        // persistently+coherently mapped storage at definition time (negative control /
+        // escape hatch). Frontend-scoped: it engages only where the active backend
+        // provides AcquirePersistentMap. With adoption on, an app SubData into a busy
+        // 128MB arena is a plain memcpy into GPU-visible memory; every driver-mediated
+        // route for the same write stalls the thread or ghost-copies the whole arena on
+        // this class of Mali driver, and the arena stops costing its size again in RAM.
+        Bool DisableLargeBufferAdoption = 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_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
+        // cross-pipeline vertex position invariance. Sorted-transparency "over" blends,
+        // gl_FragDepth writers, and fully color-masked attachments are exempt (see
+        // PipelineFactory::ShouldSuppressDepthWrite). Auto detects Qualcomm.
+        QuirkOverride MagmaDisableBlendedDepthWriteQuirk = QuirkOverride::Auto;
+        // MOBILEGL_MAGMA_DISABLE_ROBUST_BUFFER_ACCESS: leave the Vulkan robustBufferAccess device
+        // 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 MagmaDisableRobustBufferAccess = 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;
+        // MOBILEGL_SHADER_CACHE: the three-level, in-memory shader translation memo
+        // (MG_Util/ShaderTranspiler/TranslationCache.h). The levels follow the GL
+        // entry points - L1c memoizes one glCompileShader's PARSE VERDICT, L1 a
+        // linked program's whole front end, L2 DirectGLES's emitted ESSL. Auto is
+        // ON; ForceOff turns ALL THREE off and makes every translation run from
+        // scratch. The escape hatch exists because a wrong cache hit is a silently
+        // miscompiled shader: if a device ever renders differently with the cache
+        // on, one run with this falsy says so.
+        QuirkOverride ShaderTranslationCache = QuirkOverride::Auto;
+        // MOBILEGL_ESPRYT_FORCE_VIEWPORT_ARRAY_EMULATION: DirectGLES' gl_ViewportIndex routing
+        // emulation - the builtin becomes a flat varying, the fragment stage gets a
+        // per-pass gate, and a routed draw is REPLAYED once per distinct viewport state
+        // with the real glViewport/glScissor/glDepthRangef set for it. Auto is ON, and
+        // it is ON even where the driver advertises GL_OES_viewport_array, because that
+        // extension only ever gave the SHADER a compilable name: MobileGL has never
+        // programmed a driver's INDEXED viewport state (SyncRenderState pushes index 0
+        // and nothing else), so on an extension-capable driver every index rasterized as
+        // index 0 exactly as it did without one. ForceOff returns to that behaviour -
+        // the pre-emulation path, extension passthrough where it exists and
+        // LowerViewportIndexPass' demote-to-a-plain-global where it does not - and is
+        // the negative control the emulation is measured against.
+        QuirkOverride EsprytViewportArrayEmulation = QuirkOverride::Auto;
+        // MOBILEGL_ESPRYT_WIDEN_PACKED16_STORAGE: DirectGLES stores GL_RGB565/GL_RGB5(A1)/GL_RGBA4
+        // images as 8-bit-per-channel ES storage (GL_RGB8/GL_RGBA8) instead of the driver's
+        // native 16-bit packed formats. Auto defers to a POST driver-bug probe
+        // (SelfTest::CopyImageMirrorsPacked16FieldOrder): some Mali drivers store SOME
+        // packed16 allocations with a MIRRORED field order (allocation-scoped and
+        // shape/context dependent - the failing 30x30x12 GL_TEXTURE_2D_ARRAYs are mirrored
+        // at every level), so glCopyImageSubData - a raw texel-block move - lands R/G/B/A
+        // reversed whenever exactly one endpoint sits in a mirrored allocation
+        // (KHR-GL4x.copy_image.functional rgb5/rgb5_a1/rgba4 x every *2d_array* pair).
+        // With no 16-bit packed ES image left there is no field order to disagree about; the
+        // client word still round-trips exactly, because the canonical shadow is already
+        // UNorm8 and an n-bit field encodes to UNorm8 and back losslessly for n <= 8.
+        // ForceOn widens on any driver (the llvmpipe suites use it to exercise the widened
+        // path); ForceOff keeps the native narrow storage even where the probe fires - the
+        // negative control that replays the corruption. Costs 2x the memory of the affected
+        // formats where it engages, which is why Auto is probe-gated rather than always-on.
+        QuirkOverride EsprytWidenPacked16Storage = QuirkOverride::Auto;
+        // MOBILEGL_MAGMA_PRIMGEN_QUERY_REROUTE: DirectVulkan's GL_PRIMITIVES_GENERATED
+        // reroute for draws made while transform feedback is INACTIVE. The stream query
+        // (VK_QUERY_TYPE_TRANSFORM_FEEDBACK_STREAM_EXT primitivesNeeded) is defined to count
+        // them, but a Mali driver - and Mesa lavapipe - answers 0 unless a capture span is
+        // open, which is exactly the shape the CTS uses to measure the tessellator, so ~29
+        // tessellation tests per tree size a capture buffer from the 0 and die on the
+        // zero-length map. Auto defers to a device probe at renderer bring-up
+        // (SelfTest::RunPrimitivesGeneratedNoXfbProbe), which measures two substitutes on
+        // the same capture-less draws and arms the best proven one: the dedicated
+        // VK_EXT_primitives_generated_query (exact semantics by definition; lavapipe passes
+        // it, rasterizer discard included), else a clipping-invocations pipeline-statistics
+        // pool (see the verdict vocabulary for its rasterizer-discard split). ForceOn pins
+        // the reroute structurally wherever a pool can exist (the arming-observable lane,
+        // immune to the probe's verdict moving), and ForceOff is the negative control that
+        // replays the driver's silence.
+        QuirkOverride MagmaPrimGenQueryReroute = QuirkOverride::Auto;
     };
     extern FeaturesTable Features;
 } // namespace MobileGL::MG_Config

MobileGL/ConfigLoader.cpp

@@ -86,6 +86,58 @@ namespace MobileGL::MG_ConfigLoader {
         return it != acceptedEnvVariablesMap->end() && IsTruthyValue(it->second);
     }
 
+    // Quirk overrides are tri-state: an unset variable keeps device auto-detection, a truthy
+    // value forces the quirk on, anything else set ("0", "false", "") forces it off.
+    inline MG_Config::QuirkOverride QueryEnvQuirkOverride(const String& key) {
+        auto it = acceptedEnvVariablesMap->find(key);
+        if (it == acceptedEnvVariablesMap->end()) {
+            return MG_Config::QuirkOverride::Auto;
+        }
+        return IsTruthyValue(it->second) ? MG_Config::QuirkOverride::ForceOn
+                                         : 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()) {
@@ -110,19 +162,51 @@ namespace MobileGL::MG_ConfigLoader {
     inline void InitFeatures() {
         auto& features = MG_Config::Features;
         features.DisableTimerQuery = QueryEnvFlag("MOBILEGL_DISABLE_TIMERQUERY");
-        features.UseAngle = QueryEnvFlag("MOBILEGL_USE_ANGLE");
+        features.EsprytEnableTextureView = QueryEnvFlag("MOBILEGL_ESPRYT_ENABLE_TEXTURE_VIEW");
+        features.EnableSpirvValidation = QueryEnvFlag("MOBILEGL_ENABLE_SPIRV_VALIDATION");
+        features.EsprytUseAngle = QueryEnvFlag("MOBILEGL_ESPRYT_USE_ANGLE");
 #if defined(MOBILEGL_TRACE_ANGLE_VARIANTS)
         QueryEnvVariable("MOBILEGL_TRACE_ANGLE_VARIANT", features.TraceAngleVariant, "");
 #endif
-        features.DisableSubgroup = QueryEnvFlag("MOBILEGL_DISABLE_SUBGROUP");
+        features.MagmaDisableSubgroup = QueryEnvFlag("MOBILEGL_MAGMA_DISABLE_SUBGROUP");
+        features.MagmaEmulateSubgroup = QueryEnvFlag("MOBILEGL_MAGMA_EMULATE_SUBGROUP");
+        features.MagmaFixIterationRPSubgroupScratch =
+            QueryEnvQuirkOverride("MOBILEGL_MAGMA_FIX_ITERATIONRP_SUBGROUP_SCRATCH");
+        features.MagmaIterationRPFixBarrier = QueryEnvFlag("MOBILEGL_MAGMA_ITERATIONRP_FIX_BARRIER");
+        features.MagmaDeriveNumSubgroups = QueryEnvQuirkOverride("MOBILEGL_MAGMA_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.EsprytAvoidSamplerMipmapMinFilter =
+            QueryEnvFlag("MOBILEGL_ESPRYT_AVOID_SAMPLER_MIPMAP_MIN_FILTER");
+        features.EsprytAvoidExplicitLodBias = QueryEnvFlag("MOBILEGL_ESPRYT_AVOID_EXPLICIT_LOD_BIAS");
+        features.EsprytUnlocatedIoBlocks = QueryEnvQuirkOverride("MOBILEGL_ESPRYT_UNLOCATED_IO_BLOCKS");
+        features.PointSizeDemotion = QueryEnvQuirkOverride("MOBILEGL_POINT_SIZE_DEMOTION");
         features.CoherentAsFlush = QueryEnvFlag("MOBILEGL_COHERENT_AS_FLUSH");
         features.TraceSkipAutodestroy = QueryEnvFlag("MOBILEGL_TRACE_SKIP_AUTODESTROY");
-        features.DisableUboRing = QueryEnvFlag("MOBILEGL_DISABLE_UBO_RING");
+        features.EsprytDisableUboRing = QueryEnvFlag("MOBILEGL_ESPRYT_DISABLE_UBO_RING");
+        features.EsprytDisableUnpackRing = QueryEnvFlag("MOBILEGL_ESPRYT_DISABLE_UNPACK_RING");
+        features.EsprytDisableUploadRing = QueryEnvFlag("MOBILEGL_ESPRYT_DISABLE_UPLOAD_RING");
+        features.EsprytDisableInvalidateFlush = QueryEnvFlag("MOBILEGL_ESPRYT_DISABLE_INVALIDATE_FLUSH");
+        features.DisableLargeBufferAdoption = QueryEnvFlag("MOBILEGL_DISABLE_LARGE_BUFFER_ADOPTION");
+        features.EsprytForceDepthStencilReadbackEmulation =
+            QueryEnvFlag("MOBILEGL_ESPRYT_FORCE_DS_READBACK_EMULATION");
         features.RelaxedSemantics = QueryEnvFlag("MOBILEGL_RELAXED_SEMANTICS");
+        features.MagmaDisableBlendedDepthWriteQuirk =
+            QueryEnvQuirkOverride("MOBILEGL_MAGMA_DISABLE_BLENDED_DEPTH_WRITE");
+        features.MagmaDisableRobustBufferAccess = QueryEnvFlag("MOBILEGL_MAGMA_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");
+        features.ShaderTranslationCache = QueryEnvQuirkOverride("MOBILEGL_SHADER_CACHE");
+        features.EsprytViewportArrayEmulation =
+            QueryEnvQuirkOverride("MOBILEGL_ESPRYT_FORCE_VIEWPORT_ARRAY_EMULATION");
+        features.EsprytWidenPacked16Storage =
+            QueryEnvQuirkOverride("MOBILEGL_ESPRYT_WIDEN_PACKED16_STORAGE");
+        features.MagmaPrimGenQueryReroute = QueryEnvQuirkOverride("MOBILEGL_MAGMA_PRIMGEN_QUERY_REROUTE");
     }
 
     inline void InitBackendType() {

MobileGL/Defines.h

@@ -9,10 +9,20 @@
 #pragma once
 
 // ============== Platform-specific definitions and macros ============== //
-#ifdef __ANDROID__
-#undef __ANDROID_API__
-#define __ANDROID_API__ 26 // force Android API level to 26 for compatibility
-#endif
+// No __ANDROID_API__ pin here on purpose. The effective API level is owned by
+// the build system (gradle minSdk 26 -> -DANDROID_PLATFORM=android-26, enforced
+// by the configure-time guard in CMakeLists.txt), not by a macro.
+//
+// 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
@@ -34,8 +44,26 @@
 #define MOBILEGL_EGL_API MOBILEGL_API
 #define MOBILEGL_CGL_API MOBILEGL_API
 #define MOBILEGL_NSOPENGL_API MOBILEGL_API
+#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
@@ -67,6 +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/GlobalObjects.cpp

@@ -14,7 +14,13 @@ namespace MobileGL {
     } // namespace MG_Config
 
     namespace MG_Backend {
-        UniquePtr<BackendObject> pActiveBackendObject;
+        // Leak-at-exit storage: the UniquePtr itself lives on the heap and is
+        // never destroyed by the runtime, so process exit runs no backend
+        // destructors (static destruction order across TUs is undefined).
+        // Deterministic teardown happens inside the EGL lifecycle instead:
+        // the last eglTerminate calls MobileGL::Destroy(), which .reset()s
+        // these singletons while the process is still healthy.
+        UniquePtr<BackendObject>& pActiveBackendObject = *new UniquePtr<BackendObject>();
         GlobalBackendFunctionsTable gBackendFunctionsTable;
     } // namespace MG_Backend
 } // namespace MobileGL

MobileGL/Includes.h

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

MobileGL/Init.cpp

@@ -9,14 +9,30 @@
 #include "Init.h"
 #include "Config.h"
 #include <MG_Backend/BackendObjects.h>
+#include <MG_Backend/DirectVulkan/DirectVulkan.h>
 #include <MG_State/GLState/Core.h>
 #include <MG_State/EGLState/Core.h>
 #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 <MG_State/GLState/ProgramState/ProgramTranslationCache.h>
+#include <MG_Util/ShaderTranspiler/TranslationCache.h>
+
+#include <atomic>
+#include <mutex>
 
 namespace MobileGL {
     namespace {
-        Bool g_isInitialized = false;
+        std::atomic<Bool> g_isInitialized = false;
+        thread_local Bool tl_initializing = false;
+
+        std::mutex& InitMutex() {
+            static std::mutex mutex;
+            return mutex;
+        }
 
         void DestroyImpl(Bool logLifecycle) {
             if (!g_isInitialized) {
@@ -26,12 +42,46 @@ 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();
+            // The two-level translation memo. Nothing in it references a glslang object -
+            // both levels hold plain bytes - so this is RSS hygiene rather than a lifetime
+            // requirement, and it is safe either side of FinalizeProcess. Stats first: an
+            // fordebug build gets one line per level saying how the run went.
+            MG_Util::ShaderTranspiler::LogShaderTranslationCacheStats();
+            MG_Util::ShaderTranspiler::ClearShaderTranslationCaches();
+            MG_State::GLState::LogProgramTranslationCacheStats();
+            MG_State::GLState::ClearProgramTranslationCache();
             MG_Backend::gBackendFunctionsTable = {};
             g_isInitialized = false;
             if (logLifecycle) {
@@ -59,45 +109,55 @@ 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");
     }
 
+    void EnsureInitialized() {
+        if (g_isInitialized.load(std::memory_order_acquire)) {
+            return;
+        }
+        // Re-entrant call while this thread is already inside Initialize()
+        // (e.g. an init step routing back through a public entry point).
+        if (tl_initializing) {
+            return;
+        }
+        const std::lock_guard<std::mutex> lock(InitMutex());
+        if (g_isInitialized.load(std::memory_order_acquire)) {
+            return;
+        }
+        tl_initializing = true;
+        Initialize();
+        tl_initializing = false;
+    }
+
     void Destroy() {
         DestroyImpl(true);
     }
 
-#if defined(__linux__) || defined(__APPLE__)
-    __attribute__((constructor)) static void AutoInit() {
-        Initialize();
-    }
-
-    __attribute__((destructor)) static void AutoDestroy() {
-        if (MG_Config::Features.TraceSkipAutodestroy) {
-            return;
-        }
-#if defined(__APPLE__)
-        // macOS injected dylibs can run destructors after logging/backend static state is already torn down.
-        return;
-#else
-        DestroyImpl(false);
-#endif
-    }
-#endif
-
-#ifdef _WIN32
-    BOOL WINAPI DllMain(HMODULE hModule, DWORD ul_reason_for_call, LPVOID lpReserved) {
-        switch (ul_reason_for_call) {
-        case DLL_PROCESS_ATTACH:
-            Initialize();
-            break;
-
-        case DLL_PROCESS_DETACH:
-            Destroy();
-            break;
-        }
-        return TRUE;
-    }
-#endif
+    // MobileGL's lifecycle is owned entirely by the host-API layers
+    // (EGL/WGL/CGL): initialization happens lazily on the first entry point
+    // via EnsureInitialized(), and full teardown happens deterministically
+    // when the last EGL display is terminated with nothing current (EGLImpl
+    // calls Destroy()). There is intentionally no backend-initializing static
+    // constructor, no static destructor, and no DllMain: the global singletons
+    // use leak-at-exit storage (see GlobalObjects.cpp), so a process that exits
+    // without eglTerminate simply leaks them to the OS instead of running
+    // backend destructors during static teardown. macOS has a lightweight
+    // dyld constructor that installs NSOpenGL dispatch hooks only; full backend
+    // initialization still enters here from the first hooked CGL context.
 } // namespace MobileGL

MobileGL/Init.h

@@ -11,6 +11,13 @@
 
 namespace MobileGL {
     void Initialize();
+    // Thread-safe, idempotent, and re-entrant wrapper around Initialize().
+    // Host layers (EGL/WGL/CGL entry points) call this lazily on first use so
+    // full backend initialization never depends on ELF/DLL static constructors,
+    // and so a fresh init can follow a full Destroy() (e.g. after the last
+    // eglTerminate). The macOS dyld bootstrap installs only lightweight
+    // NSOpenGL method hooks.
+    void EnsureInitialized();
     void Destroy();
 
     namespace MG_Util::Debug {

MobileGL/MG_Backend/BackendObject.h

@@ -14,6 +14,7 @@ namespace MobileGL {
     namespace MG_State::GLState {
         class FramebufferObject;
         class ITextureObject;
+        class RenderbufferObject;
     }
 
     enum class BackendType {
@@ -24,6 +25,19 @@ namespace MobileGL {
     };
 
     namespace MG_Backend {
+        // One endpoint of a glCopyImageSubData. GL 4.6 core 18.3.2 accepts GL_RENDERBUFFER
+        // alongside the ten whole-image texture targets, and a renderbuffer name lives in a
+        // namespace of its own - so an endpoint is a sum type, not an ITextureObject. At most
+        // one of the two pointers is set; neither is set when the name named nothing, which is
+        // the INVALID_VALUE the frontend validator reports.
+        struct CopyImageEndpoint {
+            SharedPtr<MG_State::GLState::ITextureObject> Texture;
+            SharedPtr<MG_State::GLState::RenderbufferObject> Renderbuffer;
+
+            Bool IsRenderbuffer() const { return Renderbuffer != nullptr; }
+            Bool Exists() const { return Texture != nullptr || Renderbuffer != nullptr; }
+        };
+
         enum class FormatCapability : Uint64 {
             Creatable = 1ull << 0,
 
@@ -145,6 +159,10 @@ namespace MobileGL {
                                             GLenum buffer, GLint drawbuffer, const GLfloat* value);
             void (*ClearNamedFramebufferfi)(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
                                             GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
+            void (*ClearNamedFramebufferiv)(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
+                                            GLenum buffer, GLint drawbuffer, const GLint* value);
+            void (*ClearNamedFramebufferuiv)(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
+                                             GLenum buffer, GLint drawbuffer, const GLuint* value);
             void (*BlitFramebuffer)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0,
                                     GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
             void (*BlitNamedFramebuffer)(const SharedPtr<MG_State::GLState::FramebufferObject>& readFramebuffer,
@@ -156,9 +174,9 @@ namespace MobileGL {
                                    GLsizei height, GLint border);
             void (*CopyTexSubImage2D)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y,
                                       GLsizei width, GLsizei height);
-            void (*CopyImageSubData)(const SharedPtr<MG_State::GLState::ITextureObject>& srcTexture,
+            void (*CopyImageSubData)(const CopyImageEndpoint& src,
                                      GLenum srcTarget, GLint srcLevel, GLint srcX, GLint srcY, GLint srcZ,
-                                     const SharedPtr<MG_State::GLState::ITextureObject>& dstTexture,
+                                     const CopyImageEndpoint& dst,
                                      GLenum dstTarget, GLint dstLevel, GLint dstX, GLint dstY, GLint dstZ,
                                      GLsizei srcWidth, GLsizei srcHeight, GLsizei srcDepth);
             void (*GenerateMipmap)(GLenum target);
@@ -177,15 +195,18 @@ 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);
-            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);
+            // The GL program interface (glGetProgramInterfaceiv / glGetProgramResource*) is NOT
+            // a backend query: it describes the program the application wrote, in the
+            // application's namespace, which neither backend program is in. It is answered
+            // entirely by MG_Impl/GLImpl/Program/ProgramInterface from the frontend reflection.
+            // Takes the block's GL NAME, not glShaderStorageBlockBinding's index. The index
+            // the application passes is the frontend interface-query enumeration's, and no
+            // backend shares that index space: DirectVulkan enumerates SPIR-V descriptor
+            // bindings and DirectGLES asks a real driver about SPIRV-Cross-generated ESSL.
+            // 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
@@ -220,6 +241,39 @@ namespace MobileGL {
             // and leave the query readable later.
             Bool (*GetQueryResult64)(BackendQueryHandle query, Bool wait, Uint64* outNanoseconds);
             void (*DeleteBackendQuery)(BackendQueryHandle query);
+            // GL_SAMPLES_PASSED occlusion queries (optional; null = unsupported,
+            // the frontend then rejects the target). Results/deletion flow through
+            // GetQueryResult64 / DeleteBackendQuery like timer queries.
+            BackendQueryHandle (*BeginOcclusionQuery)();
+            void (*EndOcclusionQuery)(BackendQueryHandle query);
+            // Transform feedback primitive queries backed by real GPU query pools
+            // (optional; null = frontend falls back to CPU accounting).
+            BackendQueryHandle (*BeginXfbPrimitivesQuery)(Bool generated);
+            void (*EndXfbPrimitivesQuery)(BackendQueryHandle query);
+            // Whether GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN should be answered from the
+            // frontend's own accounting wherever that accounting is exact - a capture with no
+            // geometry stage - instead of from the query above. Set by DirectGLES, whose result
+            // is whatever the ES driver's PRIMITIVES_WRITTEN counter says: Adreno reports twice
+            // the written count for a vertex-only capture that follows a large render pass,
+            // where the desktop-exact answer is the one the frontend already computed. Defaults
+            // to false, so a backend that never sets it keeps using its GPU result.
+            Bool PrefersCpuXfbPrimitiveAccounting = false;
+            // Transform feedback capture spans, for backends whose own GL/ES driver
+            // performs the capture (DirectGLES). Both optional; null means the backend
+            // drives capture from its draw recording instead (DirectVulkan). End is
+            // called while the frontend capture state is still active, so the backend
+            // can still see the capture program and buffer bindings.
+            // GL_PATCH_VERTICES; ES 3.2 spells it the same way.
+            void (*PatchParameteri)(GLenum pname, GLint value);
+            void (*BeginTransformFeedback)(GLenum primitiveMode);
+            void (*EndTransformFeedback)();
+            // ARB_transform_feedback2. A backend that leaves these null keeps the single
+            // implicit capture span the frontend has always modelled; the frontend state
+            // (paused flag, per-object bindings) is tracked either way.
+            void (*PauseTransformFeedback)();
+            void (*ResumeTransformFeedback)();
+            void (*BindTransformFeedback)(GLuint name);
+            void (*DeleteTransformFeedback)(GLuint name);
             Int64 (*GetGpuTimestampNs)(); // glGetInteger64v(GL_TIMESTAMP); 0 if unsupported
         };
         struct GlobalBackendFunctionsTable {
@@ -230,8 +284,29 @@ namespace MobileGL {
             void (*SetSwapInterval)(Int interval);
         };
 
+        // Coarse GPU vendor identity for gating device-specific quirks. Detected from the
+        // Vulkan physical-device vendorID or the GLES GL_VENDOR/GL_RENDERER strings; stays
+        // Unknown when detection is inconclusive, in which case auto-gated quirks stay off.
+        enum class GpuVendorKind : Uint8 {
+            Unknown = 0,
+            Qualcomm,
+            Arm,
+            Nvidia,
+            Amd,
+            Intel,
+            ImgTec,
+            // Software rasterizers (llvmpipe/lavapipe, SwiftShader).
+            Software,
+        };
+
         struct DynamicBackendParameters {
             SizeT UniformBufferOffsetAlignment = 256;
+            // GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT, which is a SEPARATE limit from the
+            // uniform one and is routinely larger: Adreno 830 reports 32 for uniform buffers and
+            // 64 for storage buffers. Answering the storage query with the uniform value let an
+            // application bind a storage range at an offset the driver cannot address, which it
+            // accepted without error and then wrote somewhere else entirely.
+            SizeT ShaderStorageBufferOffsetAlignment = 256;
             // GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT. 1.0 means the backend cannot filter anisotropically,
             // which is also why the extension is not advertised in that case.
             Float MaxTextureMaxAnisotropy = 1.0f;
@@ -257,6 +332,13 @@ namespace MobileGL {
             Int MaxIntegerSamples = 1;
             Int MaxSamples = 1;
             Int MaxSampleMaskWords = 1;
+            // Tessellation limits; defaults are the GL 4.0 core minimums.
+            Int MaxPatchVertices = 32;
+            Int MaxTessGenLevel = 64;
+            // GL_MIN/MAX_PROGRAM_TEXTURE_GATHER_OFFSET. Defaults are the GL 4.0 core
+            // minimums, which every ES 3.1 driver also guarantees.
+            Int MinProgramTextureGatherOffset = -8;
+            Int MaxProgramTextureGatherOffset = 7;
             Int MaxTextureImageUnits = 32;
             Int MaxVertexTextureImageUnits = 32;
             Int MaxComputeTextureImageUnits = 32;
@@ -264,10 +346,28 @@ namespace MobileGL {
             Int MaxVertexAttribs = 16;
             Int MaxComputeShaderStorageBlocks = 8;
             Int MaxCombinedShaderStorageBlocks = 32;
+            // Per-stage GL_MAX_*_SHADER_STORAGE_BLOCKS. Zero is a legal answer for the four
+            // non-compute, non-fragment stages and these defaults are the spec minimums, not
+            // placeholders: GL 4.6 table 23.64 and ES 3.2 table 21.44 both set the minimum for
+            // vertex, tessellation control, tessellation evaluation and geometry at 0, and only
+            // fragment (8 in GL, 4 in ES) and compute are guaranteed to have any. Every real ARM
+            // GLES driver takes that allowance - a Mali-G925 reports 0 for all four - so a
+            // backend that cannot honour a graphics-stage storage block MUST report 0 here
+            // rather than a hopeful number. Advertising a non-zero count the driver will refuse
+            // does not make the block work; it only moves the failure from an honest
+            // "unsupported" at query time to a backend link error the frontend never surfaces,
+            // after which every draw with that program silently renders nothing.
+            Int MaxVertexShaderStorageBlocks = 0;
+            Int MaxTessControlShaderStorageBlocks = 0;
+            Int MaxTessEvaluationShaderStorageBlocks = 0;
+            Int MaxGeometryShaderStorageBlocks = 0;
+            Int MaxFragmentShaderStorageBlocks = 8;
             Int MaxComputeUniformBlocks = 12;
             Int MaxComputeWorkGroupInvocations = 128;
             Int MaxShaderStorageBufferBindings = 8;
             Int MaxTextureBufferSize = 65536;
+            // GL_TEXTURE_BUFFER_OFFSET_ALIGNMENT; 1 means the offset is unconstrained.
+            Int TextureBufferOffsetAlignment = 1;
             Int MaxUniformBufferBindings = 24;
             Int MaxUniformBlockSize = 16384;
             Int MaxImageUnits = 8;
@@ -278,26 +378,155 @@ namespace MobileGL {
             Int MaxComputeImageUniforms = 8;
             Int MaxDrawBuffers = 8;
             Int MaxColorAttachments = 8;
+            // GL_MAX_CLIP_DISTANCES. Zero is a legal answer here, not a placeholder, and a
+            // backend that cannot host a clip distance MUST report it: advertising eight the
+            // backend will refuse does not make gl_ClipDistance work, it only moves the failure
+            // from an honest "unsupported" at query time to a backend shader-compile error the
+            // frontend never surfaces, after which every draw with that program silently renders
+            // nothing. DirectGLES fills it from GL_EXT_clip_cull_distance, DirectVulkan from the
+            // shaderClipDistance device feature. The DEFAULT stays at the GL 4.3 core minimum
+            // because it describes the no-backend case (standalone shader compiles, unit tests),
+            // where there is no device to be honest about and BuildTBuiltInResource still has to
+            // hand glslang a workable gl_MaxClipDistances.
             Int MaxClipDistances = 8;
+            // GL_MAX_CULL_DISTANCES and GL_MAX_COMBINED_CLIP_AND_CULL_DISTANCES, under exactly
+            // the contract stated for MaxClipDistances above: ZERO IS A LEGAL ANSWER and a
+            // backend that cannot host a cull distance MUST report it. The failure this prevents
+            // is worse than the clip one, because cull distance discards the whole primitive:
+            // glslang bounds gl_CullDistance[i] against maxCullDistances and expands
+            // gl_MaxCullDistances from it, SPIRV-Cross then emits
+            // `#extension GL_EXT_clip_cull_distance : require` into the ESSL, and a host driver
+            // without that extension rejects the program in an info log nobody surfaces. These
+            // used to be bare 8s inside BuildTBuiltInResource with no backend consulted at all.
+            // The DEFAULTS are the GL 4.5 core minimums for the same reason MaxClipDistances'
+            // is: they describe the no-backend case (standalone compiles, unit tests).
+            Int MaxCullDistances = 8;
+            Int MaxCombinedClipAndCullDistances = 8;
             Int MaxViewports = 16;
+            // GL_LAYER_PROVOKING_VERTEX / GL_VIEWPORT_INDEX_PROVOKING_VERTEX: which vertex of a
+            // primitive supplies gl_Layer and gl_ViewportIndex. GL 4.6 table 23.65 makes
+            // GL_UNDEFINED_VERTEX a legal answer for both, and it is the honest default - naming
+            // a convention is a statement about behaviour, so a backend that does not pin one
+            // must not claim it does. DirectGLES fills the layer one from the ES 3.2 query and
+            // the viewport one from GL_OES_viewport_array, and leaves UNDEFINED where the
+            // capability is absent: without the viewport array extension only viewport 0 is ever
+            // rasterized, so no convention selects anything. DirectVulkan keeps UNDEFINED for
+            // both - which vertex provokes is decided per pipeline by
+            // VulkanRenderer::SelectProvokingVertexMode out of VK_EXT_provoking_vertex,
+            // provokingVertexModePerPipeline and the topology, so no single convention is true
+            // of the backend.
+            GLenum LayerProvokingVertex = GL_UNDEFINED_VERTEX;
+            GLenum ViewportIndexProvokingVertex = GL_UNDEFINED_VERTEX;
             Int MaxViewportWidth = 16384;
             Int MaxViewportHeight = 16384;
             Float ViewportBoundsRangeMin = 0.0f;
             Float ViewportBoundsRangeMax = 0.0f;
             Int ViewportSubpixelBits = 0;
+            // GL 4.x fragment-interpolation offset limits. These defaults are the
+            // core minimums and are replaced by live GLES/Vulkan device limits.
+            Float MinFragmentInterpolationOffset = -0.5f;
+            // For four fractional bits the greatest required legal offset is
+            // 0.5 - 2^-4 = 0.4375 (GL 4.6 table 23.70).
+            Float MaxFragmentInterpolationOffset = 0.4375f;
+            Int FragmentInterpolationOffsetBits = 4;
             Bool SupportsWideLines = false;
+            // Whether a framebuffer whose depth and stencil attachments are distinct
+            // images can be rendered to. GL only requires support when both refer to the
+            // same image and lets an implementation answer GL_FRAMEBUFFER_UNSUPPORTED
+            // otherwise, which is what DirectVulkan (one combined attachment) and the
+            // real ES drivers behind DirectGLES both do. Defaults to true so a backend
+            // that never sets it keeps the permissive behaviour.
+            Bool SupportsDistinctDepthStencilAttachments = true;
+            // Whether attaching a single layer of a 3D or array texture to a framebuffer actually
+            // renders to that layer. DirectGLES hands the layer straight to
+            // 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.
+            // 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 this backend can CONSUME a shader module that still declares 64-bit floats,
+            // i.e. whether `double` survives the transpile instead of being narrowed to `float`
+            // (ShaderTranspiler::DemoteFloat64Pass). Detected, never assumed:
+            //   * DirectVulkan sets it from VkPhysicalDeviceFeatures::shaderFloat64, the feature
+            //     VUID-VkShaderModuleCreateInfo-pCode-08740 requires before a module declaring
+            //     OpCapability Float64 may be created at all. lavapipe has it; Adreno and Mali
+            //     both report VK_FALSE, so no real mobile device does.
+            //   * DirectGLES can NEVER have it. GLSL ES has no 64-bit float type in any version
+            //     or extension, so SPIRV-Cross cannot emit one ("FP64 not supported in ES
+            //     profile") and the demotion there is mathematically mandatory, always.
+            // Defaults to false so a backend that never sets it - and the no-backend case, which
+            // is what standalone shader compiles and the unit tests run under - keeps the
+            // demotion, which is the behaviour that works everywhere.
+            Bool SupportsShaderFloat64 = false;
+            // 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.
+            //
+            // INDEPENDENT of SupportsShaderFloat64, and it has to be: this flag decides a VkFormat
+            // from the VAO ATTRIBUTE alone, which does not know what type the shader declared, and
+            // glVertexAttribFormat(GL_DOUBLE) feeding a plain `in vec4` is both legal and common
+            // (KHR-GL43.vertex_attrib_binding.basic-input-case4/5, advanced-bindingUpdate). A
+            // backend with native fp64 that still cannot FETCH 64 bits keeps this false and relies
+            // on the per-MODULE rule in ShaderCompiler::SanitizeAndOptimizeBinary instead: a vertex
+            // module that declares a 64-bit float INPUT is demoted whole, so the two shader-side
+            // halves (PackDoubleVertexInputsPass and VertexInputStateFactory::ToVkVertexFormat)
+            // still see one consistent world.
+            Bool SupportsFloat64VertexAttributes = false;
+            // Whether a TESSELLATION stage of this backend may access gl_PointSize - i.e.
+            // whether a module declaring OpCapability TessellationPointSize can reach the
+            // driver at all. DirectVulkan sets both this and the geometry twin from the one
+            // shaderTessellationAndGeometryPointSize feature; DirectGLES sets them
+            // independently from the EXT/OES_tessellation_point_size /
+            // geometry_point_size extension pairs (PointSizeTier), which really do come
+            // separately. When absent, ProgramSpirvTask demotes the built-in to an ordinary
+            // varying program-wide (ShaderCompiler::
+            // DemoteTessellationGeometryPointSizeForProgram); MOBILEGL_POINT_SIZE_DEMOTION
+            // overrides the detection in either direction at backend init.
+            //
+            // Defaults TRUE, deliberately against the house "assume absent" rule: false
+            // ARMS a rewrite, so the conservative no-backend answer (standalone compiles,
+            // unit tests) is the one that leaves modules untouched. A backend that never
+            // sets it gets standard modules and, at worst, the old honest declines.
+            Bool SupportsTessellationPointSize = true;
+            // The geometry-stage twin (OpCapability GeometryPointSize).
+            Bool SupportsGeometryPointSize = true;
             SizeT MaxShaderStorageBlockSize = 128 * 1024 * 1024;
             Uint32 SubgroupSize = 0;
             Uint32 SubgroupSupportedStages = 0;
             Uint32 SubgroupSupportedFeatures = 0;
             Bool SubgroupQuadOperationsInAllStages = false;
+            GpuVendorKind GpuVendor = GpuVendorKind::Unknown;
         };
 
         enum class WindowBackend {
             Android,
             X11,
             MetalLayer,
-            // TODO: Wayland, Windows, etc.
+            Win32, // Handle is an HWND
+            // TODO: Wayland, etc.
             WindowBackendCount,
             Unknown = -1
         };

MobileGL/MG_Backend/BackendObjects.h

@@ -13,6 +13,6 @@
 #include "DirectVulkan/BackendObject_DirectVulkan.h"
 
 namespace MobileGL::MG_Backend {
-    extern UniquePtr<BackendObject> pActiveBackendObject;
+    extern UniquePtr<BackendObject>& pActiveBackendObject;
     extern GlobalBackendFunctionsTable gBackendFunctionsTable;
 } // namespace MobileGL::MG_Backend

MobileGL/MG_Backend/DirectGLES/BackendObject_DirectGLES.h

@@ -18,6 +18,16 @@ namespace MobileGL::MG_Backend::DirectGLES {
                                     const MG_External::GLESCapabilities& capabilities,
                                     FormatCapabilityCache& cache);
 
+    // Clamps a requested sample count down to what the ES driver can really deliver for this
+    // format on this format-capability target: the probed per-format list when there is one, the
+    // driver's per-class GL_MAX_*_SAMPLES otherwise. The frontend deliberately validates against
+    // the count MobileGL advertises instead (GL_Getter's GetAdvertisedMaxSamples), which on a
+    // driver reporting GL_MAX_INTEGER_SAMPLES 1 is higher than the driver accepts, so every ES
+    // allocation call has to come through here. The shadow state keeps the requested count, so
+    // GL_TEXTURE_SAMPLES and framebuffer completeness still answer what the application asked for.
+    Int ClampSamplesToBackendSupport(SizeT targetIndex, TextureInternalFormat logicalFormat, GLenum imageFormat,
+                                     Int samples);
+
     class BackendObject_DirectGLES : public BackendObject {
     public:
         ~BackendObject_DirectGLES() override;
@@ -67,9 +77,13 @@ 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 / EXT-OES texture views 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,
+                                                  Bool textureViewSupported, Bool cubeMapArraySupported);
 
     // 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);
@@ -59,6 +61,10 @@ namespace MobileGL::MG_Backend::DirectGLES {
                                  GLenum buffer, GLint drawbuffer, const GLfloat* value);
     void ClearNamedFramebufferfi(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
                                  GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
+    void ClearNamedFramebufferiv(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
+                                 GLenum buffer, GLint drawbuffer, const GLint* value);
+    void ClearNamedFramebufferuiv(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
+                                  GLenum buffer, GLint drawbuffer, const GLuint* value);
     void BlitFramebuffer(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1,
                          GLint dstY1, GLbitfield mask, GLenum filter);
     void BlitNamedFramebuffer(const SharedPtr<MG_State::GLState::FramebufferObject>& readFramebuffer,
@@ -70,9 +76,9 @@ namespace MobileGL::MG_Backend::DirectGLES {
                         GLsizei height, GLint border);
     void CopyTexSubImage2D(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width,
                            GLsizei height);
-    void CopyImageSubData(const SharedPtr<MG_State::GLState::ITextureObject>& srcTexture,
+    void CopyImageSubData(const CopyImageEndpoint& src,
                           GLenum srcTarget, GLint srcLevel, GLint srcX, GLint srcY, GLint srcZ,
-                          const SharedPtr<MG_State::GLState::ITextureObject>& dstTexture,
+                          const CopyImageEndpoint& dst,
                           GLenum dstTarget, GLint dstLevel, GLint dstX, GLint dstY, GLint dstZ,
                           GLsizei srcWidth, GLsizei srcHeight, GLsizei srcDepth);
     void GenerateMipmap(GLenum target);
@@ -88,15 +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);
-    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 ShaderStorageBlockBinding(GLuint program, const GLchar* storageBlockName, GLuint storageBlockBinding);
     Bool InitWindowSurface(NativeWindowType window);
     Bool InitPbufferSurface(EGLint width, EGLint height);
     Bool MakeCurrent();
@@ -121,6 +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
@@ -130,6 +146,16 @@ namespace MobileGL::MG_Backend::DirectGLES {
     BackendQueryHandle BeginTimeElapsedQuery();
     void EndTimeElapsedQuery(BackendQueryHandle query);
     BackendQueryHandle QueryCounterTimestamp();
+    // GL_ANY_SAMPLES_PASSED(_CONSERVATIVE) occlusion queries: core ES3, independent of
+    // GL_EXT_disjoint_timer_query and of MOBILEGL_DISABLE_TIMERQUERY. Results/deletion
+    // flow through GetQueryResult64/DeleteBackendQuery like the timer queries above.
+    BackendQueryHandle BeginOcclusionQuery();
+    void EndOcclusionQuery(BackendQueryHandle query);
+    // GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN / GL_PRIMITIVES_GENERATED, also core ES
+    // (GL_PRIMITIVES_GENERATED from ES 3.2 on). Null when the target is unavailable, in
+    // which case the frontend falls back to counting primitives from the draw calls.
+    BackendQueryHandle BeginXfbPrimitivesQuery(Bool generated);
+    void EndXfbPrimitivesQuery(BackendQueryHandle query);
     Bool IsQueryResultAvailable(BackendQueryHandle query);
     // Returns true when a final value landed in *outNanoseconds (a zero for
     // null or stale-generation handles IS final: the frontend may cache it
@@ -146,6 +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);
@@ -154,6 +185,49 @@ namespace MobileGL::MG_Backend::DirectGLES {
     void SetGLESCapabilities(const MG_External::GLESCapabilities& capabilities);
     void DestroyEGLContext();
 
+    // Transform feedback capture spans, performed by the real ES driver. The
+    // capture set is declared on the backend program at link time; the driver-side
+    // begin is deferred to the first draw of the span (ES needs the capturing
+    // program current and the capture buffers bound), and the end also mirrors the
+    // captured bytes back into the frontend buffer shadows.
+    void PatchParameteri(GLenum pname, GLint value);
+
+    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();
+        void ResumeTransformFeedback();
+        void BindTransformFeedback(GLuint name);
+        void DeleteTransformFeedback(GLuint name);
+        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/MultiDraw.cpp

@@ -0,0 +1,972 @@
+// 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 index value this batch restarts on, compared at 32 bits against the zero-extended
+        // source index. Normally the all-ones value of the source type, which is what
+        // GL_PRIMITIVE_RESTART_FIXED_INDEX and GLES both restart on; with desktop
+        // GL_PRIMITIVE_RESTART it is instead whatever glPrimitiveRestartIndex named. The rebased
+        // tier turns whichever it is into 0xFFFFFFFF in its widened stream, which is what the
+        // driver restarts on.
+        //
+        // No truncation, deliberately, and the same rule ResolveRestartSubstitution applies: a
+        // restart index the source type cannot hold simply matches nothing, so returning it
+        // verbatim is already "this batch restarts nowhere".
+        Uint32 RestartSentinelFor(GLenum type) {
+            if (ResolveRestartSubstitution(type) != RestartSubstitutionKind::None) {
+                return MG_State::pGLContext->GetPrimitiveRestartIndex();
+            }
+            return MG_Util::FixedRestartIndexForGLType(type);
+        }
+
+        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, Bool arbitraryRestart) {
+            ResolveTierOnce();
+            GLESMultiDrawMode tier = g_resolvedTier;
+
+            // Desktop GL_PRIMITIVE_RESTART restarts on an application-chosen index; the driver
+            // only ever restarts on the all-ones value. Every tier but the rebased one hands
+            // the application's own index data to the driver, which would then see no restarts
+            // at all and weld the primitives together. The rebased tier is the one that
+            // REWRITES the stream, and RestartSentinelFor already tells it which value to
+            // translate, so it is the only tier this batch can take.
+            if (arbitraryRestart) {
+                return GLESMultiDrawMode::DrawElements;
+            }
+
+            // 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) {
+                ForEachViewportRoutingPass([&] {
+                    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);
+                    ForEachViewportRoutingPass([&] {
+                        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);
+                ForEachViewportRoutingPass([&] {
+                    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);
+            // Widening to GL_UNSIGNED_INT gives a UBYTE/USHORT source a sentinel it can never
+            // spell, so those batches are lossless. A UINT source that already uses 0xFFFFFFFF as
+            // a real vertex index while restarting on a different one is the one shape 32 bits
+            // cannot express - the same corner the single-draw substitution reports.
+            if (restartActive && indexSize == 4 && restartSentinel != 0xFFFFFFFFu) {
+                MGLOG_E_ONCE("GL_PRIMITIVE_RESTART with restart index %u over GL_UNSIGNED_INT multi-draw indices: "
+                             "any index that is already 0xFFFFFFFF will restart too, because the rewritten stream "
+                             "has no wider sentinel to move to.",
+                             restartSentinel);
+            }
+            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);
+                ForEachViewportRoutingPass([&] {
+                    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;
+        // Read before any GL work, because it decides the tier below: a desktop restart index
+        // the driver does not know about can only be honoured by the tier that rewrites the
+        // index stream (see ResolveTierForBatch). A restart index this index type cannot hold
+        // needs no rewrite at all - nothing can match it - but it does need the driver's own
+        // fixed-index restart held off for the batch, which is what the scope below does.
+        const RestartSubstitutionKind restartKind = ResolveRestartSubstitution(type);
+        const Bool arbitraryRestart = restartKind == RestartSubstitutionKind::RewriteIndices;
+        const ScopedSuppressedPrimitiveRestart restartCapOverride(restartKind);
+
+        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);
+            ForEachViewportRoutingPass([&] {
+                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, arbitraryRestart);
+
+        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 - except that the base-vertex replay hands the
+        // application's own indices to the driver, which cannot restart on a desktop
+        // restart index, so that batch has only the rewriting floor.
+        if (!drawn && !arbitraryRestart) {
+            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.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 {
@@ -34,13 +36,148 @@ 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);
+
+        // Whether this format's ES storage is widened to 8-bit-per-channel because the
+        // driver stores some packed16 allocations with a mirrored field order
+        // (PixelFormatNormalizeOptionBit::WidenPacked16Norm). True only for
+        // GL_RGB565/GL_RGB5(_A1)/GL_RGBA4, and only where the POST probe measured the
+        // divergence (or MOBILEGL_ESPRYT_WIDEN_PACKED16_STORAGE forces it). The transfer paths
+        // consult it too: the packed-norm re-upload leg must stand down when the ES storage
+        // is no longer 16-bit packed.
+        Bool UsesWidenedPacked16NormStorage(TextureInternalFormat internalFormat);
+
         void GenerateTextureFormatInfo(TextureInternalFormat internalFormat, GLenum* outInternalFormat,
                                        GLenum* outFormat, GLenum* outType,
                                        TextureTarget target = TextureTarget::Unknown);
         void GenerateRenderbufferFormatInfo(TextureInternalFormat internalFormat, GLenum* outInternalFormat,
                                             GLenum* outFormat, GLenum* outType);
         Bool ShouldUseCaveatTextureFormat(TextureInternalFormat internalFormat, TextureTarget target);
+
+        // True when the format the image is actually created with has an alpha channel the
+        // frontend format does not (the three-channel colour-renderable widening). GL reads such
+        // 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);
+
+        // The CHANNEL WIDENING an image-bindable texture's ES storage takes, so that a format
+        // GLSL ES cannot spell as an image is carried by one it can.
+        //
+        // GL has forty image formats, GLSL ES core has thirteen, and no test device advertises
+        // GL_NV_image_formats - so a shader declaring one of the other twenty-six has no legal
+        // ESSL at all and glBindImageTexture rejects the narrow format outright for most of them
+        // (GL_INVALID_VALUE for nineteen of twenty-six on Adreno, twenty-five on both Malis).
+        // Seventeen have a core format of the SAME per-channel width and component type,
+        // differing only in channel count, and in one of those the emulation is EXACT: GL already
+        // defines an imageLoad from a narrower format as (r, 0, 0, 1) and an imageStore as
+        // dropping the components the format does not have, so the carrier's surplus channels
+        // hold values GL has already named. WidenImageFormatsPass pins them in the shader; this
+        // is the storage half, and DirectGLES::TextureImpl::SyncImageTextureBinding the bind
+        // half. All three ask WidenedCoreEsslImageFormat, so they cannot pick different carriers.
+        //
+        // Reports nothing (InternalFormat == GL_UNKNOWN_MGL) for a format that is core already,
+        // for the nine with no exact carrier (r11f_g11f_b10f, rgb10_a2, rgb10_a2ui, rgba16, rg16,
+        // r16, rgba16_snorm, rg16_snorm, r16_snorm - those keep the honest "no GLSL ES spelling"
+        // diagnostic rather than a silent approximation), and on a driver that HAS
+        // GL_NV_image_formats, where the shader keeps the declared format and no widening may
+        // happen behind it.
+        //
+        // The widened triple REPLACES what GenerateTextureFormatInfo chose, including any
+        // renderability substitution: an image that cannot be image-bound is useless whatever its
+        // attachment behaviour, so the image constraint wins. In practice that only bites
+        // RG8_SNORM/R8_SNORM on a driver without EXT_render_snorm, where the storage stays
+        // signed-normalized instead of becoming the half float that fallback would have picked -
+        // so an image-bound texture in one of those two formats is no longer attachable, and
+        // glGetTexImage on it falls through to the CPU shadow, which a shader-side imageStore
+        // does not update. Accepted deliberately: before the widening, an image binding in either
+        // format was refused outright by every driver tested and the stage that declared it never
+        // compiled at all, so nothing that works today is being given up.
+        //
+        // KNOWN GAP, for the same "all three layers move together" reason: a widened texture that
+        // is ALSO an FBO colour attachment gains one to three writable channels, and a draw into
+        // it can leave values in channels GL says are 0 and 1. Sampling and imageLoad are covered
+        // (the swizzle composition in SyncTextureParamsToBackend and the shader-side mask), but a
+        // glReadPixels/glGetTexImage that asks for more channels than the frontend format has
+        // would see them. Closing it needs the per-draw-buffer colour mask the three-channel
+        // widening already carries (FramebufferImpl::g_alphaWidenedDrawBufferMask) generalized
+        // from "alpha" to a channel count, which is its own change.
+        // How the FRONTEND's CPU shadow for a widened format is laid out relative to the carrier's
+        // transfer, i.e. what the upload has to do to it. Almost every entry is `Components`: the
+        // shadow already holds SourceChannels components of exactly the carrier's own type, so
+        // padding it out to four is the whole conversion. The packed entries do not - their shadow
+        // is ONE 32-bit word per texel - and reading such a word as components of the carrier's
+        // type takes twelve or sixteen bytes out of four and shears the level.
+        enum class ImageWidenSourceEncoding : Uint8 {
+            Components = 0,
+            // r11f_g11f_b10f: GL_UNSIGNED_INT_10F_11F_11F_REV -> four GL_FLOATs of an rgba16f.
+            PackedFloat11f11f10f,
+            // rgb10_a2 and rgb10_a2ui: GL_UNSIGNED_INT_2_10_10_10_REV -> four GL_UNSIGNED_SHORT
+            // channel CODES of an rgba16ui. The same split serves both: the two formats differ
+            // only in what the codes MEAN, which is the shader's business and not the transfer's.
+            PackedInt2101010Rev,
+        };
+
+        struct ImageBindableStorageWidening {
+            GLenum InternalFormat = GL_UNKNOWN_MGL;
+            GLenum Format = GL_UNKNOWN_MGL;
+            GLenum Type = GL_UNKNOWN_MGL;
+            // Channels the FRONTEND format has, i.e. how many of the carrier's four the client
+            // data fills. The rest are uploaded as 0, and the fourth as the format's implied 1.
+            Uint SourceChannels = 0;
+            // Whether that implied 1 is the integer one or a saturated normalized field - the
+            // transfer type cannot tell the two apart (GL_UNSIGNED_BYTE serves both RG8 and
+            // RG8UI), so the carrier decides.
+            Bool IntegerData = false;
+            // What the upload has to do to the frontend shadow before it describes the level to
+            // the driver (PrepareImageWidenedUpload).
+            ImageWidenSourceEncoding SourceEncoding = ImageWidenSourceEncoding::Components;
+            // Non-zero when the carrier holds this format's channels as the INTEGER CODES of a
+            // NORMALIZED value - the seven 16-bit and 10-bit normalized formats, which core ESSL
+            // has no image format of any width for and which a float carrier would requantise.
+            // Each entry is the largest code that channel can hold, i.e. the denominator of GL 4.6
+            // 2.3.5; SignedNormalized picks which of the two conversions it is the denominator of.
+            //
+            // Two things depend on it, both because the ES storage no longer shares the frontend
+            // format's component class: the upload pads a missing alpha with ChannelMax[3] instead
+            // of the transfer type's own "one" (through a uint carrier the saturated field IS the
+            // one), and glGetTexImage divides the codes back out into the floats the application
+            // is still owed.
+            Uint ChannelMax[4] = {0u, 0u, 0u, 0u};
+            Bool SignedNormalized = false;
+
+            Bool CarriesNormalizedCodes() const { return ChannelMax[0] != 0u; }
+            explicit operator Bool() const { return InternalFormat != GL_UNKNOWN_MGL; }
+        };
+        ImageBindableStorageWidening GetImageBindableStorageWidening(TextureInternalFormat internalFormat);
+
+        // The single-channel core format an image-bindable BUFFER texture's view is SPLIT into, or
+        // GL_UNKNOWN_MGL for a format that needs no split (or has no core base).
+        //
+        // A buffer texture cannot be widened: its texels are the application's buffer object, at
+        // the size and layout the application gave it, and it is usually also a vertex, index or
+        // storage buffer whose bytes are not ours to restride. But an rg32f view of N texels and
+        // an r32f view of 2N texels describe exactly the SAME bytes, so the split changes only
+        // how the shader subscripts them - component j of texel i is texel 2i + j of the base
+        // view - which WidenImageFormatsPass rewrites every access to do. The same rule as the
+        // widening decides WHETHER: a driver that can spell rg32f for an imageBuffer needs
+        // nothing.
+        //
+        // KNOWN GAP, and the reason this is not applied to a texture that is merely sampled: a
+        // buffer texture that is BOTH image-bound and read through a samplerBuffer would have its
+        // sampled view split too, and the sampler side is not rewritten. Accepted for the same
+        // reason the storage widening's gaps are - on a driver where the split applies at all
+        // there is no legal ESSL for the image declaration, so such a program did not compile.
+        GLenum GetImageBindableBufferSplitFormat(TextureInternalFormat internalFormat);
     } // namespace TextureImpl
 
     namespace FramebufferImpl {} // namespace FramebufferImpl
@@ -88,6 +225,24 @@ namespace MobileGL::MG_Backend::DirectGLES {
         // bytes, dst receives width * GetReadbackDstPixelSize(mapping, type) bytes.
         void ConvertWideReadbackRow(const Uint8* src, Uint8* dst, SizeT width, GLenum wideType,
                                     const ReadbackChannelMapping& mapping, GLenum type);
+
+        // Stores wide RGBA(_INTEGER) rows into the client pointer or the bound PACK pixel buffer,
+        // honoring the client-side PACK pixel-store parameters (row length, alignment, skips,
+        // swap-bytes, and - when applyPackImageParams - image height/skip images). Shared by the
+        // DirectGLES and DirectVulkan readback conversion paths.
+        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 {
@@ -96,7 +251,289 @@ namespace MobileGL::MG_Backend::DirectGLES {
         String ClampNormFallbackOutputs(String glslCode, GLenum shaderType, Uint32 snormOutputMask,
                                         Uint32 unormOutputMask);
         String ForceFlatIntegerVaryings(const String& glslCode, GLenum shaderType);
+        // Legacy GLSL's gl_FragColor is broadcast to every enabled draw buffer (GL 4.6
+        // 15.2.3), but ShaderSourceProcessor lowers it to the single output mg_FragColor,
+        // which only ever reaches draw buffer 0. Replicates it across `drawBufferCount`
+        // outputs and copies the value into them at the end of main. A no-op for
+        // 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);
+        // Adds `#extension GL_OES_viewport_array : require` when the emitted ESSL names
+        // gl_ViewportIndex. SPIRV-Cross prints that identifier and asks for nothing (unlike
+        // gl_Layer, which it backs with GL_NV_viewport_array2 on ES) and ESSL has no core
+        // spelling for it at any version, so the request has to be made here or the stage does
+        // not compile - which loses the whole program, not just the multi-viewport routing.
+        // `needed` is the caller's answer for the same reason as above: only it knows whether the
+        // driver advertises the extension, and requesting an unadvertised one is itself a compile
+        // error, so this is never emitted speculatively. A no-op when not needed or already
+        // present.
+        String RequestViewportArrayExtension(String glslCode, Bool needed);
+        // Adds `#extension <extensionName> : require` when a TESSELLATION or GEOMETRY stage's
+        // emitted ESSL names gl_PointSize. Desktop GL has that built-in in gl_PerVertex for every
+        // vertex-processing stage; ESSL does NOT have it in those two at any version - not even
+        // 320, where the stages themselves are core - until EXT/OES_tessellation_point_size resp.
+        // EXT/OES_geometry_point_size is requested. SPIRV-Cross prints the identifier bare and
+        // asks for nothing, exactly as it does for gl_ViewportIndex, so without this the stage
+        // fails to compile with "`gl_PointSize' undeclared" and the WHOLE program is replaced by
+        // program 0 - the draw renders nothing and any transform-feedback capture it was carrying
+        // is rejected outright. `extensionName` is the caller's answer, nullptr when the driver
+        // advertises neither spelling, because requesting an unadvertised extension is itself a
+        // compile error. A no-op when nullptr or already present.
+        String RequestPointSizeExtension(String glslCode, const char* extensionName);
+        // The extension name RequestPointSizeExtension should be given for `tier`, or nullptr for
+        // PointSizeTier::None. `tessellation` picks the tessellation spellings over the geometry
+        // ones; the two extensions are separate and neither implies the other.
+        const char* PointSizeExtensionName(MG_External::GLESCapabilities::PointSizeTier tier, Bool tessellation);
+        // 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 per-element scalar declarations RemapImageArrayElementUnits splits an
+        // image array into; the suffix is the array's own name and the element's index.
+        constexpr const char* IMAGE_ARRAY_ELEMENT_PREFIX = "mg_imageElem_";
+        // One image ARRAY whose elements the application pointed at units that are not
+        // consecutive-from-element-zero.
+        struct ImageArrayUnitPlan {
+            String name;       // the array's name, exactly as the emitted ESSL declares it
+            Vector<Int> units; // the frontend image unit element k has to reach
+        };
+        // Desktop GL lets an application give each element of an image array an ARBITRARY unit
+        // (glUniform1i per element). ES has no such call at all - "ES image units come
+        // exclusively from the layout(binding=N) qualifier" - and one declaration carries one
+        // binding, so ESSL nails an array's elements to the CONSECUTIVE units N, N+1, N+2, ...
+        // MobileGL used to stamp element [0]'s unit as the binding and let the rest fall where
+        // they fell: KHR-GL4x.shader_image_load_store.advanced-sso-simple assigns 0,2,4,6 and
+        // 1,3,5,7, so its two programs actually addressed 0,1,2,3 and 1,2,3,4 - one layer got the
+        // wrong value and three were never written, with no GL error and no link log. The same
+        // defect for SAMPLER arrays was fixed API-side (SubscriptUniformNameForElement); an image
+        // array has no API side to fix, because ES makes glUniform1i on an image uniform an
+        // INVALID_OPERATION.
+        //
+        // Repaired by SPLITTING the array into one SCALAR image uniform per element, each with
+        // its own layout(binding = N), and rewriting `name[k]` to the scalar declared for
+        // element k. One declaration carries one binding, so one declaration per unit is the
+        // only spelling that reaches an arbitrary set of them.
+        //
+        // That rewrite needs every k in the emitted text to be a LITERAL, and it is:
+        // LegalizeResourceArrayIndexingForEssl has already folded or lowered every dynamic
+        // image-array subscript in the module, because ESSL forbids one outright ("image arrays
+        // indexed with non-constant expressions are forbidden in GLSL ES", Mesa 26.1.4 at
+        // ES 3.2, on a raw GLES probe with no MobileGL in the loop). The earlier shape here -
+        // widening the array to cover the whole span of units and routing each subscript through
+        // a `const highp int` offset table - was written before that pass covered images, and
+        // the table lookup was itself one of the non-constant expressions the same probe refuses.
+        // The split also costs exactly the image uniforms the application declared, where the
+        // widening cost the whole SPAN (seven for the four elements of
+        // KHR-GL42.shader_image_load_store.advanced-sso-simple), so there is no budget for it to
+        // fail to fit in.
+        //
+        // Declines - leaving the array exactly as it was, and naming it in `outDeclined` for the
+        // caller to report - when the emitted extent disagrees with the reflection, when the
+        // array is reached by anything other than a subscript, or when a subscript is not a
+        // literal element index. Silence was the whole defect here, so a decline must be audible.
+        //
+        // Must run AFTER RebindImageUniformsToFrontendUnits and BakeImageFormatQualifiers (both
+        // key on the GL uniform name and on a binding already being stamped) and BEFORE
+        // SplitReadWriteImageUniforms (so each element that is both read and written is split
+        // with its own binding already on it) and RemoveLayoutBinding (which is what preserves
+        // image bindings). Like them, it is downstream of the L2 shader-translation memo, so the
+        // per-program units it reads need no entry in BuildEsslTranslationKey.
+        String RemapImageArrayElementUnits(const String& glslCode, const Vector<ImageArrayUnitPlan>& plans,
+                                           Vector<String>* outDeclined = nullptr);
+        // The member list of a `gl_PerVertex { ... }` redeclaration in already-emitted ESSL -
+        // the text between the braces, verbatim - or nullopt when the shader does not redeclare
+        // the block in that direction. `input` selects the `in gl_PerVertex` form over the
+        // `out` one.
+        //
+        // Exists so BuildPassthroughTessControlEssl can MIRROR the stages it has to sit between
+        // rather than guess at them. Whether SPIRV-Cross redeclares the built-in block, and with
+        // which members, depends on what the application's shader touched; a synthesized stage
+        // that redeclares a different shape than its neighbours is an ES link error against a
+        // program that has no other problem.
+        std::optional<String> ExtractPerVertexBlockMembers(const String& essl, Bool input);
+        // The pass-through tessellation control stage GL 4.6 core 11.2.2 describes: "the input
+        // patch is passed through unmodified", the output patch has PATCH_VERTICES vertices, and
+        // the levels come from the PATCH_DEFAULT_OUTER_LEVEL / PATCH_DEFAULT_INNER_LEVEL state.
+        //
+        // Desktop GL makes the control stage OPTIONAL. OpenGL ES 3.2 does not: it has no
+        // PATCH_DEFAULT_*_LEVEL state at all (only glPatchParameteri, for PATCH_VERTICES) and
+        // rejects a program that has an evaluation stage without a control stage - with an EMPTY
+        // info log, verified on an Adreno 830 with no MobileGL in the process. MobileGL's own
+        // frontend link succeeds, so the program reports GL_LINK_STATUS = TRUE, program 0 is
+        // bound in its place, and every draw silently renders nothing.
+        //
+        // `inPerVertexMembers` / `outPerVertexMembers` are the member lists to redeclare gl_in
+        // and gl_out with - normally taken from the neighbouring stages' own emitted ESSL via
+        // ExtractPerVertexBlockMembers, and empty to leave the driver's built-in declaration
+        // alone, which is what matching a neighbour that did not redeclare requires.
+        //
+        // All four outer levels and both inner levels are written unconditionally: writing a
+        // level the evaluation stage's domain does not use is legal and ignored, and it saves
+        // this from having to know the domain. They are the GL_PATCH_DEFAULT_OUTER_LEVEL /
+        // GL_PATCH_DEFAULT_INNER_LEVEL state, baked in as literals - ES has no such state and no
+        // glPatchParameterfv to forward to, so compiling them in is the only way to honour them.
+        // That makes them part of what a built program is stale against, exactly as PATCH_VERTICES
+        // is: see the staleness clause in DirectGLES.cpp's SyncCurrentProgram, which compares both.
+        //
+        // The same stage, for the same reason, that DirectVulkan synthesizes in
+        // ProgramFactory::BuildPassthroughTessControlSource - Vulkan likewise requires both
+        // tessellation stages. Kept as two generators rather than one because the two targets
+        // disagree on everything but the algorithm: desktop GLSL 450 against ESSL, a fixed
+        // gl_PerVertex shape that Vulkan matches structurally against a mirrored one, and a
+        // VkShaderModule against a driver shader object.
+        String BuildPassthroughTessControlEssl(Uint esslVersion, Uint patchVertices,
+                                               const String& inPerVertexMembers,
+                                               const String& outPerVertexMembers,
+                                               const FloatVec4& defaultOuterLevel,
+                                               const FloatVec2& defaultInnerLevel);
+        // Prefix of the writeonly half a read+write image uniform is split into (see
+        // SplitReadWriteImageUniforms); the suffix is the image's own (already access-tagged) name.
+        constexpr const char* IMAGE_WRITE_ALIAS_PREFIX = "mg_imageWrite_";
+        // The three names SplitReadWriteImageUniforms renames a rewritten image declaration
+        // under, one per REPAIR it can apply. Which one a stage picks is decided by that stage's
+        // own accesses, so two stages that use an image the same way arrive at the SAME name and
+        // two that use it differently arrive at different ones - which is exactly the property
+        // the rename exists for, at no cost to the stages that agree. Exposed for the tests.
+        constexpr const char* IMAGE_READONLY_ALIAS_PREFIX = "mg_imageRo_";
+        constexpr const char* IMAGE_WRITEONLY_ALIAS_PREFIX = "mg_imageWo_";
+        constexpr const char* IMAGE_SPLIT_READ_ALIAS_PREFIX = "mg_imageRw_";
+        // 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, and RENAMES it after the repair it applied while doing so:
+        //  * loaded only            -> add `readonly`, rename under IMAGE_READONLY_ALIAS_PREFIX
+        //  * stored only            -> add `writeonly`, rename under IMAGE_WRITEONLY_ALIAS_PREFIX
+        //  * both                   -> emit TWO declarations on the same binding and of the
+        //                              same type, `coherent readonly
+        //                              <IMAGE_SPLIT_READ_ALIAS_PREFIX><name>` and `coherent
+        //                              writeonly <IMAGE_WRITE_ALIAS_PREFIX><that name>`, point
+        //                              every imageStore at the second one, and follow each of
+        //                              those stores with `memoryBarrierImage();`. 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.
+        //
+        // The rename is the other half of the repair and applies to all three cases. The qualifier
+        // chosen above is a decision about ONE STAGE's accesses, and GLSL requires a uniform
+        // declared in two stages to be declared identically - so a shader that stores an image from
+        // the vertex stage and loads it from the fragment stage came out of here `writeonly` in one
+        // and `readonly` in the other. Adreno merges the two same-named declarations and silently
+        // drops the vertex-stage STORES: no GL error, no link log, LINK_STATUS = 1, and the image
+        // still reads back its initial contents
+        // (KHR-GL4x.shader_image_load_store.advanced-memory-dependentInvocation; a raw-ES probe
+        // isolated the trigger to the same-name/mismatched-qualifier pair, and only when both
+        // carry `coherent`). Renaming leaves no cross-stage variable to merge.
+        //
+        // The name is keyed on the REPAIR, not on the stage, and that distinction is the whole
+        // point: two stages that use an image the same way emit byte-identical declarations, so
+        // letting them keep one shared name costs nothing and merging them is correct, while two
+        // stages that use it differently land on different prefixes and cannot be merged at all.
+        // A per-STAGE tag also satisfied the first requirement but violated the second: it made
+        // the SAME image a distinct uniform in every stage that named it, and Adreno allocates
+        // image LOCATIONS per distinct uniform. KHR-GL43.shading_language_420pack.
+        // binding_images_texture_type_* declares three read+write images in each of its five
+        // stages; merged that is 6 image uniforms, per-stage-tagged it is 30, and the Adreno 830
+        // linker answered "Error: Image Image location or component exceeds max allowed. Error:
+        // Linking failed." - which, the frontend having already published LINK_STATUS = TRUE from
+        // glslang's link, surfaced only as every draw silently doing nothing and the images
+        // reading back zero. Mali and Mesa link the same text, so nothing but a device gate
+        // catches this.
+        //
+        // A declaration SPIRV-Cross already tagged `readonly` or `writeonly` needs no qualifier
+        // repair, but it is NOT stage-independent: that tag is derived from the accesses of the
+        // stage being emitted, so an image stored in the vertex stage and loaded in the fragment
+        // stage arrives here as `coherent writeonly g_image` and `coherent readonly g_image` -
+        // one name, two spellings, which is exactly the pair Adreno merges. Those declarations
+        // are therefore renamed too, keyed on the qualifier they already carry (readonly ->
+        // IMAGE_READONLY_ALIAS_PREFIX, writeonly -> IMAGE_WRITEONLY_ALIAS_PREFIX) and with
+        // nothing but the identifier changed. Stages that agree still reach the same alias and
+        // stay merged, so this costs no shader an extra image uniform.
+        //
+        // The declarations this pass still leaves untouched keep their names: one carrying BOTH
+        // readonly and writeonly (a spelling no access analysis produces, so it came from the
+        // application and is identical everywhere), and one carrying NEITHER, which is legal only
+        // for the r32f/r32i/r32ui formats and is likewise spelled the same in every stage.
+        //
+        // The `coherent` on both halves of the pair is load-bearing, not decoration: GLSL only
+        // guarantees a write through one image variable is visible to a read through a DIFFERENT
+        // one when both are coherent, and the split is what makes a same-variable
+        // read-after-write cross-variable. The single-declaration repairs above do not get it -
+        // nothing aliases them.
+        //
+        // The barrier is the other half of the same problem, and coherent alone did not cover it:
+        // visibility is not ORDER. Within one invocation the ES compiler sees a write to one
+        // variable and a read of another it has no reason to believe alias, and is free to serve
+        // the read from before the write - which is what advanced-memory-order's store/load/
+        // compare loop measured on Adreno. memoryBarrierImage() orders exactly those two, is core
+        // GLSL ES 3.10 in every stage, and is not an execution barrier, so it is legal in
+        // non-uniform control flow. It costs something in a shader that stores to a read+write
+        // image in a loop, which is why it is confined to the split pair.
+        //
+        // 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. Its output is a function of the emitted text alone - it needs no
+        // stage and no per-program state - so it adds nothing to BuildEsslTranslationKey either.
+        //
+        // `outSplitCount`, when given, receives the number of declarations that were actually
+        // doubled - i.e. exactly how many image uniforms this stage gained over what the
+        // application declared. Zero for every shader but a handful, and the only number the
+        // budget note above can be reported with.
+        String SplitReadWriteImageUniforms(const String& glslCode, Uint* outSplitCount = nullptr);
+        // 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_";
+        // ES has no per-texture/sampler LOD bias at all (GL_TEXTURE_LOD_BIAS is desktop
+        // only; Vulkan spells it VkSamplerCreateInfo::mipLodBias), so it has to reach the
+        // shader as a uniform and be folded into every lookup's level of detail. Declares
+        // one `uniform highp float mg_lodBias_<sampler>;` per mip-capable sampler and adds
+        // it to the bias / explicit-LOD argument of every lookup that takes one. Draws push
+        // 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.
+        //
+        // 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_ESPRYT_AVOID_EXPLICIT_LOD_BIAS).
+        String EmulateTextureLodBias(const String& glslCode, Bool avoidExplicitLodBias = false);
     } // namespace PrgramImpl
 
     namespace Utils {

MobileGL/MG_Backend/DirectVulkan/BackendObject_DirectVulkan.cpp

@@ -9,15 +9,19 @@
 #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>
 #include <cstdlib>
 #include <cstring>
 
@@ -40,13 +44,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Bool IsLayeredTarget(TextureTarget target) {
             return target == TextureTarget::Texture3D || target == TextureTarget::Texture1DArray ||
                    target == TextureTarget::Texture2DArray || target == TextureTarget::TextureCubeMap ||
-                   target == TextureTarget::TextureCubeMapArray ||
-                   target == TextureTarget::Texture2DMultisampleArray;
+                   target == TextureTarget::TextureCubeMapArray || target == TextureTarget::Texture2DMultisampleArray;
         }
 
         Bool IsMultisampleTarget(TextureTarget target) {
-            return target == TextureTarget::Texture2DMultisample ||
-                   target == TextureTarget::Texture2DMultisampleArray;
+            return target == TextureTarget::Texture2DMultisample || target == TextureTarget::Texture2DMultisampleArray;
         }
 
         Bool IsTextureBufferTarget(TextureTarget target) {
@@ -58,8 +60,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             GLenum normalizedInternalFormat = glFormat;
             GLenum imageFormat = GL_RGBA;
             GLenum imageType = GL_UNSIGNED_BYTE;
-            MG_Util::TextureFormatProcessor::NormalizePixelFormat(
-                glFormat, PixelFormatNormalizeOptionBit::None, &normalizedInternalFormat, &imageFormat, &imageType);
+            MG_Util::TextureFormatProcessor::NormalizePixelFormat(glFormat, PixelFormatNormalizeOptionBit::None,
+                                                                  &normalizedInternalFormat, &imageFormat, &imageType);
             return imageFormat == GL_RED_INTEGER || imageFormat == GL_RG_INTEGER || imageFormat == GL_RGB_INTEGER ||
                    imageFormat == GL_RGBA_INTEGER;
         }
@@ -80,8 +82,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return caps;
         }
 
-        FormatCapabilityFlags BuildVulkanCaps(TextureInternalFormat logicalFormat,
-                                              TextureTarget target,
+        FormatCapabilityFlags BuildVulkanCaps(TextureInternalFormat logicalFormat, TextureTarget target,
                                               VkFormatFeatureFlags features) {
             FormatCapabilityFlags caps;
             const Bool isDepth = MG_Util::IsDepthFormatInternalFormat(logicalFormat);
@@ -100,8 +101,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             const Bool sampled = (features & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT) != 0;
             const Bool linearFilter = (features & VK_FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR_BIT) != 0;
             const Bool colorRenderable = (features & VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT) != 0;
-            const Bool depthStencilRenderable =
-                (features & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT) != 0;
+            const Bool depthStencilRenderable = (features & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT) != 0;
             const Bool renderable = (isDepth || isStencil) ? depthStencilRenderable : colorRenderable;
 
             if (sampled || renderable) {
@@ -140,6 +140,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             case TextureInternalFormat::RGB:
             case TextureInternalFormat::RGB8:
                 return TextureInternalFormat::RGBA8;
+            // Legacy low-bit-depth formats with no (or rarely supported) native Vulkan
+            // encoding; a wider normalized fallback keeps at least the required precision.
+            case TextureInternalFormat::R3G3B2:
+            case TextureInternalFormat::RGB4:
+            case TextureInternalFormat::RGB5:
+            case TextureInternalFormat::RGBA2:
+            case TextureInternalFormat::RGBA4:
+            case TextureInternalFormat::RGB5A1:
+                return TextureInternalFormat::RGBA8;
+            case TextureInternalFormat::RGB10:
+                return TextureInternalFormat::RGB10A2;
+            case TextureInternalFormat::RGB12:
+            case TextureInternalFormat::RGBA12:
+                return TextureInternalFormat::RGBA16;
             case TextureInternalFormat::SRGB8:
                 return TextureInternalFormat::SRGB8Alpha8;
             case TextureInternalFormat::RGB8Snorm:
@@ -184,21 +198,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         Bool HasNewCaveatFormatCaps(FormatCapabilityFlags nativeCaps, FormatCapabilityFlags fallbackCaps) {
             for (FormatCapability capability : kReportedFormatCapabilities) {
-                if (HasFormatCapability(fallbackCaps, capability) &&
-                    !HasFormatCapability(nativeCaps, capability)) {
+                if (HasFormatCapability(fallbackCaps, capability) && !HasFormatCapability(nativeCaps, capability)) {
                     return true;
                 }
             }
             return false;
         }
 
-        void LogVulkanFormatCaveat(TextureInternalFormat logicalFormat,
-                                   SizeT targetIndex,
+        void LogVulkanFormatCaveat(TextureInternalFormat logicalFormat, SizeT targetIndex,
                                    TextureInternalFormat fallbackFormat) {
-            MGLOG_D("Caveat: %s %s not fully supported. Reason: native Vulkan format is not fully supported. Fallback: %s",
-                    GetFormatCapabilityTargetName(targetIndex).c_str(),
-                    MG_Util::ConvertTextureInternalFormatToString(logicalFormat).c_str(),
-                    MG_Util::ConvertTextureInternalFormatToString(fallbackFormat).c_str());
+            MGLOG_D(
+                "Caveat: %s %s not fully supported. Reason: native Vulkan format is not fully supported. Fallback: %s",
+                GetFormatCapabilityTargetName(targetIndex).c_str(),
+                MG_Util::ConvertTextureInternalFormatToString(logicalFormat).c_str(),
+                MG_Util::ConvertTextureInternalFormatToString(fallbackFormat).c_str());
         }
 
         Vector<Int> BuildSampleCounts(Int maxSamples) {
@@ -242,15 +255,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
                 for (SizeT targetIndex = 0; targetIndex < kFormatCapabilityTextureTargetCount; ++targetIndex) {
                     const auto target = static_cast<TextureTarget>(targetIndex);
-                    const VkFormatFeatureFlags nativeFeatures =
-                        IsTextureBufferTarget(target) ? nativeProperties.bufferFeatures
-                                                      : nativeProperties.optimalTilingFeatures;
+                    const VkFormatFeatureFlags nativeFeatures = IsTextureBufferTarget(target)
+                                                                    ? nativeProperties.bufferFeatures
+                                                                    : nativeProperties.optimalTilingFeatures;
                     FormatCapabilityFlags nativeCaps = BuildVulkanCaps(logicalFormat, target, nativeFeatures);
                     cache.FullCaps[targetIndex][formatIndex] |= nativeCaps;
 
-                    const VkFormatFeatureFlags fallbackFeatures =
-                        IsTextureBufferTarget(target) ? fallbackProperties.bufferFeatures
-                                                      : fallbackProperties.optimalTilingFeatures;
+                    const VkFormatFeatureFlags fallbackFeatures = IsTextureBufferTarget(target)
+                                                                      ? fallbackProperties.bufferFeatures
+                                                                      : fallbackProperties.optimalTilingFeatures;
                     FormatCapabilityFlags fallbackCaps = BuildVulkanCaps(logicalFormat, target, fallbackFeatures);
                     if (fallbackFormat != VK_FORMAT_UNDEFINED && fallbackFormat != nativeFormat) {
                         cache.CaveatCaps[targetIndex][formatIndex] |= fallbackCaps;
@@ -285,9 +298,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 cache.FullCaps[renderbufferTargetIndex][formatIndex] |= renderbufferCaps;
 
                 if (fallbackFormat != VK_FORMAT_UNDEFINED && fallbackFormat != nativeFormat) {
-                    FormatCapabilityFlags fallbackRenderbufferCaps =
-                        BuildVulkanCaps(logicalFormat, TextureTarget::Texture2D,
-                                        fallbackProperties.optimalTilingFeatures);
+                    FormatCapabilityFlags fallbackRenderbufferCaps = BuildVulkanCaps(
+                        logicalFormat, TextureTarget::Texture2D, fallbackProperties.optimalTilingFeatures);
                     fallbackRenderbufferCaps &= FormatCapability::Creatable;
                     if ((fallbackProperties.optimalTilingFeatures &
                          (VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT | VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) !=
@@ -296,8 +308,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                         fallbackRenderbufferCaps |= FormatCapability::MultisampleRenderbuffer;
                     }
                     cache.CaveatCaps[renderbufferTargetIndex][formatIndex] |= fallbackRenderbufferCaps;
-                    if (fallbackLogicalFormat &&
-                        HasNewCaveatFormatCaps(renderbufferCaps, fallbackRenderbufferCaps)) {
+                    if (fallbackLogicalFormat && HasNewCaveatFormatCaps(renderbufferCaps, fallbackRenderbufferCaps)) {
                         LogVulkanFormatCaveat(logicalFormat, renderbufferTargetIndex, *fallbackLogicalFormat);
                     }
                 }
@@ -314,14 +325,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
     void PopulateFormatCapabilities(VkPhysicalDevice physicalDevice,
                                     PFN_vkGetPhysicalDeviceFormatProperties getFormatProperties,
-                                    const MG_External::VulkanCapabilities& capabilities,
-                                    FormatCapabilityCache& cache) {
+                                    const MG_External::VulkanCapabilities& capabilities, FormatCapabilityCache& cache) {
         PopulateFormatCapabilitiesImpl(physicalDevice, getFormatProperties, capabilities, cache);
     }
 
     BackendObject_DirectVulkan::~BackendObject_DirectVulkan() = default;
 
-    BackendObject_DirectVulkan::BackendObject_DirectVulkan(): m_rendererInfo{GetRendererIdentity()} {}
+    BackendObject_DirectVulkan::BackendObject_DirectVulkan() : m_rendererInfo{GetRendererIdentity()} {}
 
     Bool BackendObject_DirectVulkan::InitWindowSurface() {
         if (!m_windowHandle.Handle) {
@@ -375,6 +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());
@@ -395,10 +408,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             MGLOG_E("DirectVulkan backend not initialized");
             return false;
         }
-        if (!handle.Handle || (handle.Backend != WindowBackend::Android &&
-                               handle.Backend != WindowBackend::X11 &&
-                               handle.Backend != WindowBackend::MetalLayer)) {
-            MGLOG_E("DirectVulkan backend only supports Android, X11, and CAMetalLayer native windows");
+        if (!handle.Handle || (handle.Backend != WindowBackend::Android && handle.Backend != WindowBackend::X11 &&
+                               handle.Backend != WindowBackend::MetalLayer && handle.Backend != WindowBackend::Win32)) {
+            MGLOG_E("DirectVulkan backend only supports Android, X11, CAMetalLayer, and Win32 native windows");
             return false;
         }
 
@@ -454,6 +466,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // treat them as signaled/available with zero results from here on.
         BumpRendererGeneration();
         pVulkanRenderer.reset();
+        // The reflection cache is file-scope, not renderer-owned; without this the
+        // deleted programs' reflection strings survive full context teardown.
+        ClearProgramResourceCaches();
         BackendObject::ReleaseEGLResources();
     }
 
@@ -463,6 +478,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // treat them as signaled/available with zero results from here on.
         BumpRendererGeneration();
         pVulkanRenderer.reset();
+        // The reflection cache is file-scope, not renderer-owned; without this the
+        // deleted programs' reflection strings survive full context teardown.
+        ClearProgramResourceCaches();
     }
 
     const RendererInfo& BackendObject_DirectVulkan::GetRendererInfo() const {
@@ -482,35 +500,158 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             .RendererName = "Magma",
             .BackendName = "Direct (Vulkan)",
             .ExtraVendor = Nullopt,
-            .RendererGLInfo =
-                {
-                    .TargetGLVersion = {3, 3, 0},
-                    .TargetGLSLVersion = {4, 6, 0},
-                    // Baseline advertisement (no shader subgroup, no timer queries); a
-                    // live backend reconciles its copy in UpdateAdvertisedExtensions.
-                    .Extensions = BuildAdvertisedExtensions(false, false, false),
-                    .IsCompatibilityProfile = false
-                },
+            .RendererGLInfo = {.TargetGLVersion = {4, 6, 0},
+                               .TargetGLSLVersion = {4, 6, 0},
+                               // Baseline advertisement (no runtime-gated capabilities); a live
+                               // backend reconciles its copy in UpdateAdvertisedExtensions.
+                               .Extensions = BuildAdvertisedExtensions(false, false, false, false, false),
+                               .IsCompatibilityProfile = false},
             .StaticBackendCapability = {.AllowVSOnlyPrograms = false}};
         return rendererInfo;
     }
 
     Vector<GLExtension> BuildAdvertisedExtensions(Bool shaderSubgroupSupported, Bool timerQueriesSupported,
-                                                  Bool anisotropicFilteringSupported) {
-        Vector<GLExtension> extensions = {V_OpenGL30, V_OpenGL31, V_OpenGL32,
-                                          V_OpenGL33, 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_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_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};
-        if (shaderSubgroupSupported && !MG_Config::Features.DisableSubgroup) {
+                                                  Bool anisotropicFilteringSupported,
+                                                  Bool nonZeroIndirectBaseInstanceSupported,
+                                                  Bool cubeMapArraySupported) {
+        Vector<GLExtension> extensions = {
+            // The version tokens have to reach the version the backend actually claims:
+            // TargetGLVersion is {4,6,0}, and a list that stopped at OpenGL40 told an
+            // application feature-detecting off these tokens the opposite of what
+            // GL_MAJOR_VERSION / GL_MINOR_VERSION told it.
+            V_OpenGL30, V_OpenGL31, V_OpenGL32, V_OpenGL33, V_OpenGL40, V_OpenGL41, V_OpenGL42, V_OpenGL43,
+            V_OpenGL44, V_OpenGL45, V_OpenGL46,
+            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_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,
+            // Unconditional, unlike DirectGLES: a GL texture view is a second set of VkImageViews
+            // over the same VkImage with a sub-range and possibly a reinterpreted VkFormat, which
+            // is core Vulkan on every device MobileGL runs on. Format-reinterpreting views need
+            // VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT on the image, which SyncTextureResource sets for
+            // every immutable-storage texture (see the comment there).
+            E_GL_ARB_texture_view,
+            // Core since 3.2 and implemented here on both backends - glDrawElementsBaseVertex,
+            // glDrawRangeElementsBaseVertex, glDrawElementsInstancedBaseVertex and
+            // glMultiDrawElementsBaseVertex all reach real per-draw vertex rebasing. The string
+            // was simply never emitted, which left KHR-GL4*.draw_elements_base_vertex_tests
+            // NotSupported on a feature that works.
+            E_GL_ARB_draw_elements_base_vertex,
+            // The whole sync-object family is real and core since 3.2: glFenceSync, glIsSync,
+            // glDeleteSync, glClientWaitSync, glWaitSync and glGetSynciv all live in GLImpl over a
+            // backend fence (a VkFence here, an EGLSync/GLsync on DirectGLES), and glGetInteger64v
+            // answers GL_MAX_SERVER_WAIT_TIMEOUT. The string matters for the same reason
+            // ARB_uniform_buffer_object's does: LWJGL builds GLCapabilities from the extension
+            // list, and a caller that finds GL_ARB_sync missing never resolves the entry points -
+            // then calls through null if it uses fences anyway. Nothing in the CTS gates on this
+            // string, so it is advertised on the strength of the implementation, not a test unlock.
+            E_GL_ARB_sync,
+            // Atomic counters, core since 4.2. glGetActiveAtomicCounterBufferiv and the whole
+            // GL_ATOMIC_COUNTER_BUFFER_* query family are real in GLImpl, and the counter buffer
+            // now reaches the shader on BOTH backends - Magma resolves the lowered
+            // gl_AtomicCounterBlock_<N> from the atomic-counter binding points rather than the
+            // shader-storage ones (see ResolveStorageBufferDescriptor). Withheld here until that
+            // landed, because the counter silently read whatever was bound as SSBO N instead.
+            E_GL_ARB_shader_atomic_counters,
+            // glVertexAttribDivisor, core since 3.3 and real on both backends. Applications
+            // (Better Clouds' GLCompat among them) accept the extension string as an
+            // ALTERNATIVE to a 3.3 context when deciding whether instanced rendering is
+            // available, so withholding it makes MobileGL look less capable than it is.
+            E_GL_ARB_instanced_arrays,
+            // Core GL 3.0-4.3 plumbing that has been real here for as long as the backend has
+            // existed, and that was simply never named. None of these unlocks a single CTS case -
+            // the conformance suite reaches all of them through the version - so they are
+            // advertised for the OTHER consumer of this list: LWJGL builds GLCapabilities from the
+            // string set, and an application that gates its ENTRY POINTS on the string rather than
+            // on the version never resolves them and then calls through null. Each is backed by
+            // the entry points named beside it. Kept identical to the DirectGLES block so the two
+            // backends do not disagree about what MobileGL is.
+            //
+            // glBindVertexArray / glGenVertexArrays / glDeleteVertexArrays / glIsVertexArray.
+            E_GL_ARB_vertex_array_object,
+            // The 14 glSamplerParameter* / glGetSamplerParameter* entry points, including the
+            // integer-valued Iiv/Iuiv forms.
+            E_GL_ARB_sampler_objects,
+            // glMapBufferRange + glFlushMappedBufferRange, which ARB_buffer_storage's persistent
+            // maps are already built on top of.
+            E_GL_ARB_map_buffer_range,
+            // glCopyBufferSubData plus the GL_COPY_READ_BUFFER / GL_COPY_WRITE_BUFFER targets.
+            E_GL_ARB_copy_buffer,
+            // glCopyImageSubData, wired to a real backend hook on both backends.
+            E_GL_ARB_copy_image,
+            // GL_TEXTURE_SWIZZLE_{R,G,B,A,RGBA}, which map onto a VkImageView's component swizzle.
+            E_GL_ARB_texture_swizzle,
+            // GL_INT_2_10_10_10_REV / GL_UNSIGNED_INT_2_10_10_10_REV on glVertexAttribPointer plus
+            // the eight glVertexAttribP* entry points.
+            E_GL_ARB_vertex_type_2_10_10_10_rev,
+            // The R/RG internal formats. Named separately from the float ones because an
+            // application may check either.
+            E_GL_ARB_texture_rg,
+            // GL_DEPTH_COMPONENT32F and GL_DEPTH32F_STENCIL8.
+            E_GL_ARB_depth_buffer_float,
+            // The floating-point colour formats. Unlike the rest of this block this string DOES
+            // gate CTS cases - KHR-GL4*.internalformat.texture2d.*{16f,32f} is keyed on it with no
+            // core-version fallback, so eight cases per version list were NotSupported on formats
+            // the backend has always had.
+            E_GL_ARB_texture_float,
+            // glViewportArrayv / glViewportIndexedf{,v} / glScissorArrayv / glScissorIndexed{,v} /
+            // glDepthRangeArrayv / glDepthRangeIndexed / glGetFloati_v / glGetDoublei_v, over the
+            // 16 viewports GL_MAX_VIEWPORTS reports.
+            E_GL_ARB_viewport_array,
+            // 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.MagmaDisableSubgroup) {
             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), and stays opt-in even on a
+        // device that HAS shaderFloat64. Every `double` in a shader compiles and runs either way
+        // - narrowed to 32 bits where the device has no 64-bit floats, kept whole where it does -
+        // so an application that simply uses doubles needs nothing advertised. What the extension
+        // additionally promises is the whole GL_ARB_gpu_shader_fp64 SURFACE (glUniform*d
+        // conformance, the fp64 built-ins, the state queries), and turning the string on is a
+        // decision about all of it rather than about the shader path alone.
+        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.
@@ -524,6 +665,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             extensions.push_back(E_GL_EXT_texture_filter_anisotropic);
             extensions.push_back(E_GL_ARB_texture_filter_anisotropic);
         }
+        // A cube map array is a 6n-layer VkImage viewed as VK_IMAGE_VIEW_TYPE_CUBE_ARRAY, and that
+        // view type cannot be created without the imageCubeArray device feature - so the string
+        // follows the feature, not the version, exactly as the per-layer attachment bit does.
+        //
+        // Named for the application's benefit rather than the suite's: measured on Adreno 830,
+        // KHR-GL43.texture_gather.plain-gather-*-cube-array already passed without the string, so
+        // this unlocks no conformance case. It is advertised because the feature is real and
+        // because an application that feature-detects cube map arrays off the string (rather than
+        // off the 4.0 version) would otherwise decline a path this backend serves.
+        if (cubeMapArraySupported) {
+            extensions.push_back(E_GL_ARB_texture_cube_map_array);
+        }
         return extensions;
     }
 
@@ -570,6 +723,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             funcsTable.GL.ClearBufferiv = ClearBufferiv;
             funcsTable.GL.ClearNamedFramebufferfv = ClearNamedFramebufferfv;
             funcsTable.GL.ClearNamedFramebufferfi = ClearNamedFramebufferfi;
+            funcsTable.GL.ClearNamedFramebufferiv = ClearNamedFramebufferiv;
+            funcsTable.GL.ClearNamedFramebufferuiv = ClearNamedFramebufferuiv;
             funcsTable.GL.BlitFramebuffer = BlitFramebuffer;
             funcsTable.GL.BlitNamedFramebuffer = BlitNamedFramebuffer;
             funcsTable.GL.CopyTexImage2D = CopyTexImage2D;
@@ -587,12 +742,6 @@ 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;
@@ -613,6 +762,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 funcsTable.GL.DeleteBackendQuery = DeleteBackendQuery;
                 funcsTable.GL.GetGpuTimestampNs = GetGpuTimestampNs;
             }
+            // Occlusion queries share the handle-based result/delete entries, which must
+            // exist even when timer queries are disabled.
+            funcsTable.GL.BeginOcclusionQuery = BeginOcclusionQuery;
+            funcsTable.GL.EndOcclusionQuery = EndOcclusionQuery;
+            funcsTable.GL.BeginXfbPrimitivesQuery = BeginXfbPrimitivesQuery;
+            funcsTable.GL.EndXfbPrimitivesQuery = EndXfbPrimitivesQuery;
+            funcsTable.GL.IsQueryResultAvailable = IsQueryResultAvailable;
+            funcsTable.GL.GetQueryResult64 = GetQueryResult64;
+            funcsTable.GL.DeleteBackendQuery = DeleteBackendQuery;
             funcsTableInitialized = true;
         }
         return funcsTable;
@@ -627,6 +785,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_vulkanCaps = capabilities;
         UpdateDynamicBackendParameters();
         UpdateAdvertisedExtensions();
+        if (MG_State::pGLContext) {
+            MG_State::pGLContext->InvalidateCompileEnv();
+        }
         MutableFormatCapabilities().Clear();
     }
 
@@ -637,9 +798,17 @@ 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(),
-            pVulkanRenderer && pVulkanRenderer->IsSamplerAnisotropySupported());
+            subgroupSupportAdvertised, pVulkanRenderer && pVulkanRenderer->IsTimerQuerySupported(),
+            pVulkanRenderer && pVulkanRenderer->IsSamplerAnisotropySupported(),
+            pVulkanRenderer && pVulkanRenderer->IsNonZeroIndirectBaseInstanceSupported(),
+            m_vulkanCaps.SupportsImageCubeArray);
     }
 
     void BackendObject_DirectVulkan::UpdateDynamicBackendParameters() {
@@ -687,13 +856,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         static constexpr SizeT kMaxAdvertisedShaderStorageBlockSize = 512ull * 1024ull * 1024ull;
         m_dynamicParameters.UniformBufferOffsetAlignment = m_vulkanCaps.UniformBufferOffsetAlignment;
+        m_dynamicParameters.ShaderStorageBufferOffsetAlignment = m_vulkanCaps.ShaderStorageBufferOffsetAlignment;
         m_dynamicParameters.AliasedLineWidthRangeMin = m_vulkanCaps.AliasedLineWidthRangeMin;
         m_dynamicParameters.AliasedLineWidthRangeMax = m_vulkanCaps.AliasedLineWidthRangeMax;
         // Without the samplerAnisotropy feature the limit is unusable, so report 1.0 (no anisotropy)
         // rather than a maximum the sampler manager will never apply.
         m_dynamicParameters.MaxTextureMaxAnisotropy =
             (pVulkanRenderer && pVulkanRenderer->IsSamplerAnisotropySupported()) ? m_vulkanCaps.MaxSamplerAnisotropy
-                                                                                : 1.0f;
+                                                                                 : 1.0f;
         m_dynamicParameters.SmoothLineWidthRangeMin = m_vulkanCaps.SmoothLineWidthRangeMin;
         m_dynamicParameters.SmoothLineWidthRangeMax = m_vulkanCaps.SmoothLineWidthRangeMax;
         m_dynamicParameters.SmoothLineWidthGranularity = m_vulkanCaps.SmoothLineWidthGranularity;
@@ -714,8 +884,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_dynamicParameters.MaxIntegerSamples = m_vulkanCaps.MaxIntegerSamples;
         m_dynamicParameters.MaxSamples = m_vulkanCaps.MaxSamples;
         m_dynamicParameters.MaxSampleMaskWords = m_vulkanCaps.MaxSampleMaskWords;
-        const Int maxSupportedTextureUnits =
-            static_cast<Int>(MG_State::GLState::TextureState::MAX_TEXTURE_IMAGE_UNITS);
+        const Int maxSupportedTextureUnits = static_cast<Int>(MG_State::GLState::TextureState::MAX_TEXTURE_IMAGE_UNITS);
         // GL_MAX_TEXTURE_IMAGE_UNITS is a *per-stage* sampler limit. Adreno/Qualcomm report a huge
         // maxPerStageDescriptorSampledImages (descriptor-indexing scale), so clamping it only to our
         // combined array capacity (192) still advertises 192 per stage. Host code treats this value as
@@ -725,8 +894,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // limits while keeping the combined limit at our texture-unit array capacity.
         constexpr Int maxPerStageTextureUnits =
             static_cast<Int>(MG_State::GLState::TextureState::MAX_PER_STAGE_TEXTURE_IMAGE_UNITS);
-        m_dynamicParameters.MaxTextureImageUnits =
-            std::min(m_vulkanCaps.MaxTextureImageUnits, maxPerStageTextureUnits);
+        m_dynamicParameters.MaxTextureImageUnits = std::min(m_vulkanCaps.MaxTextureImageUnits, maxPerStageTextureUnits);
         m_dynamicParameters.MaxVertexTextureImageUnits =
             std::min(m_vulkanCaps.MaxVertexTextureImageUnits, maxPerStageTextureUnits);
         m_dynamicParameters.MaxComputeTextureImageUnits =
@@ -735,19 +903,84 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             std::min(m_vulkanCaps.MaxCombinedTextureImageUnits, maxSupportedTextureUnits);
         // Never advertise more attributes than the state layer can store: the current-value array and
         // 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;
-        m_dynamicParameters.MaxCombinedShaderStorageBlocks = m_vulkanCaps.MaxCombinedShaderStorageBlocks;
-        m_dynamicParameters.MaxComputeUniformBlocks = m_vulkanCaps.MaxComputeUniformBlocks;
+        m_dynamicParameters.MaxVertexAttribs = std::min(
+            m_vulkanCaps.MaxVertexAttribs, static_cast<Int>(MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS));
+        // Vulkan descriptor limits are not GL limits, and a GL application reads an advertised
+        // limit as an amount it may actually USE. Adreno answers the per-stage/per-set descriptor
+        // 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.MaxTextureBufferSize = m_vulkanCaps.MaxTextureBufferSize;
-        m_dynamicParameters.MaxUniformBufferBindings = m_vulkanCaps.MaxUniformBufferBindings;
+        m_dynamicParameters.MaxShaderStorageBufferBindings =
+            clampLimit("GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS", m_vulkanCaps.MaxShaderStorageBufferBindings,
+                       kMaxAdvertisedBufferBlocks);
+        // Per-stage GL_MAX_*_SHADER_STORAGE_BLOCKS. Vulkan has one descriptor limit for every
+        // stage (maxPerStageDescriptorStorageBuffers, which is what MaxComputeShaderStorageBlocks
+        // carries), so the stage limits differ only by whether the stage can have blocks at all.
+        //
+        // Deliberately NOT gated on vertexPipelineStoresAndAtomics, unlike the per-stage image
+        // uniforms below. That gate reads as the obvious one and is wrong here in practice: a
+        // Mali-G925-Immortalis reports vertexPipelineStoresAndAtomics=false (supported AND
+        // enabled) and yet runs all 433 KHR-GL43.constant_expressions.*_tess_* cases correctly
+        // through this backend - those write their result through a storage block declared in a
+        // tessellation stage. Gating would report 0 and turn 433 passing cases into
+        // "unsupported", removing function that demonstrably works.
+        //
+        // The asymmetry with DirectGLES is real and is the point. There, 0 prevents a program
+        // the driver refuses outright at link time; the honest limit converts a silent
+        // wrong-render into a capability an application can route around. Here there is no such
+        // failure to prevent, so the limit stays at what the device can address. If a Vulkan
+        // device is ever found that genuinely rejects such a pipeline, the gate belongs at
+        // pipeline creation where the rejection is observable, not on a feature bit this driver
+        // reports inaccurately.
+        {
+            const Int maxPerStageStorageBlocks =
+                std::min(std::max(m_dynamicParameters.MaxComputeShaderStorageBlocks, 0),
+                         std::min(std::max(m_dynamicParameters.MaxCombinedShaderStorageBlocks, 0),
+                                  std::max(m_dynamicParameters.MaxShaderStorageBufferBindings, 0)));
+            m_dynamicParameters.MaxVertexShaderStorageBlocks = maxPerStageStorageBlocks;
+            m_dynamicParameters.MaxTessControlShaderStorageBlocks = maxPerStageStorageBlocks;
+            m_dynamicParameters.MaxTessEvaluationShaderStorageBlocks = maxPerStageStorageBlocks;
+            // The one hard capability in the set: no geometry stage means no blocks in it.
+            m_dynamicParameters.MaxGeometryShaderStorageBlocks =
+                m_vulkanCaps.SupportsGeometryShader ? maxPerStageStorageBlocks : 0;
+            m_dynamicParameters.MaxFragmentShaderStorageBlocks = maxPerStageStorageBlocks;
+        }
+        m_dynamicParameters.MaxTextureBufferSize = clampLimit(
+            "GL_MAX_TEXTURE_BUFFER_SIZE", m_vulkanCaps.MaxTextureBufferSize, kMaxAdvertisedTextureBufferSize);
+        m_dynamicParameters.TextureBufferOffsetAlignment = m_vulkanCaps.TextureBufferOffsetAlignment;
+        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.MaxImageUnits = std::max(std::min(m_vulkanCaps.MaxImageUnits, maxSupportedTextureUnits), 0);
         m_dynamicParameters.MaxCombinedImageUniforms = std::max(m_vulkanCaps.MaxCombinedImageUniforms, 0);
         const Int maxPerStageImageUniforms =
             std::min(m_dynamicParameters.MaxImageUnits, m_dynamicParameters.MaxCombinedImageUniforms);
@@ -764,25 +997,156 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             m_vulkanCaps.SupportsFragmentStoresAndAtomics ? maxPerStageImageUniforms : 0;
         m_dynamicParameters.MaxComputeImageUniforms =
             std::min(std::max(m_vulkanCaps.MaxComputeImageUniforms, 0), maxPerStageImageUniforms);
-        const Int maxSupportedDrawBuffers =
-            static_cast<Int>(MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS);
+        const Int maxSupportedDrawBuffers = static_cast<Int>(MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS);
         m_dynamicParameters.MaxDrawBuffers = std::min(m_vulkanCaps.MaxDrawBuffers, maxSupportedDrawBuffers);
         m_dynamicParameters.MaxColorAttachments = std::min(m_vulkanCaps.MaxColorAttachments, maxSupportedDrawBuffers);
-        m_dynamicParameters.MaxClipDistances = m_vulkanCaps.MaxClipDistances;
+        // Same shape as the image-uniform limits three lines above: maxClipDistances is reported
+        // by every device, but declaring ClipDistance in a module needs the shaderClipDistance
+        // FEATURE, which VulkanRenderer enables exactly where the physical device has it. Without
+        // it the limit describes a capacity no shader may use, so report none.
+        m_dynamicParameters.MaxClipDistances =
+            m_vulkanCaps.SupportsShaderClipDistance ? std::max(m_vulkanCaps.MaxClipDistances, 0) : 0;
+        // The cull pair, gated on its own feature. shaderCullDistance is separate from
+        // shaderClipDistance and VulkanRenderer enables it independently, so it gets its own
+        // gate rather than riding on the clip one.
+        m_dynamicParameters.MaxCullDistances =
+            m_vulkanCaps.SupportsShaderCullDistance ? std::max(m_vulkanCaps.MaxCullDistances, 0) : 0;
+        // GL 4.6 core 11.1.3.10: the combined limit is at least as large as either half. A device
+        // with only one of the two features must not report a combined capacity that implies the
+        // other, so the gate is "either feature" and the value never drops below what is enabled.
+        m_dynamicParameters.MaxCombinedClipAndCullDistances =
+            (m_vulkanCaps.SupportsShaderClipDistance || m_vulkanCaps.SupportsShaderCullDistance)
+                ? std::max({m_vulkanCaps.MaxCombinedClipAndCullDistances, m_dynamicParameters.MaxClipDistances,
+                            m_dynamicParameters.MaxCullDistances})
+                : 0;
         m_dynamicParameters.MaxViewports = m_vulkanCaps.MaxViewports;
+        // Assigned explicitly rather than left to the struct's defaults, like every other
+        // parameter here, so a second fill cannot inherit a stale value. GL_UNDEFINED_VERTEX is
+        // the truthful answer for DirectVulkan and a legal one (GL 4.6 table 23.65): which vertex
+        // provokes is chosen per pipeline by VulkanRenderer::SelectProvokingVertexMode out of
+        // VK_EXT_provoking_vertex, provokingVertexModePerPipeline and the topology, so there is no
+        // one convention to name. Vulkan's own default is FIRST, which is the opposite of the
+        // GL_LAST_VERTEX_CONVENTION this used to claim unconditionally.
+        m_dynamicParameters.LayerProvokingVertex = GL_UNDEFINED_VERTEX;
+        m_dynamicParameters.ViewportIndexProvokingVertex = GL_UNDEFINED_VERTEX;
         m_dynamicParameters.MaxViewportWidth = m_vulkanCaps.MaxViewportWidth;
         m_dynamicParameters.MaxViewportHeight = m_vulkanCaps.MaxViewportHeight;
         m_dynamicParameters.ViewportBoundsRangeMin = m_vulkanCaps.ViewportBoundsRangeMin;
         m_dynamicParameters.ViewportBoundsRangeMax = m_vulkanCaps.ViewportBoundsRangeMax;
         m_dynamicParameters.ViewportSubpixelBits = m_vulkanCaps.ViewportSubpixelBits;
+        m_dynamicParameters.MinFragmentInterpolationOffset =
+            std::isfinite(m_vulkanCaps.MinFragmentInterpolationOffset) &&
+                    m_vulkanCaps.MinFragmentInterpolationOffset <= -0.5f
+                ? m_vulkanCaps.MinFragmentInterpolationOffset
+                : -0.5f;
+        m_dynamicParameters.MaxFragmentInterpolationOffset = 0.4375f;
+        m_dynamicParameters.FragmentInterpolationOffsetBits = 4;
+        if (m_vulkanCaps.FragmentInterpolationOffsetBits >= 4 &&
+            std::isfinite(m_vulkanCaps.MaxFragmentInterpolationOffset)) {
+            const Float requiredMaxOffset = 0.5f - std::ldexp(1.0f, -m_vulkanCaps.FragmentInterpolationOffsetBits);
+            if (m_vulkanCaps.MaxFragmentInterpolationOffset >= requiredMaxOffset) {
+                m_dynamicParameters.MaxFragmentInterpolationOffset = m_vulkanCaps.MaxFragmentInterpolationOffset;
+                m_dynamicParameters.FragmentInterpolationOffsetBits = m_vulkanCaps.FragmentInterpolationOffsetBits;
+            }
+        }
         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);
+            }
+        }
+        // The device feature the whole fp64 story hangs off. With it, a module keeps its
+        // OpCapability Float64 and real doubles reach the driver; without it the transpile
+        // narrows every 64-bit float to 32 (ShaderTranspiler::DemoteFloat64Pass), because
+        // VUID-VkShaderModuleCreateInfo-pCode-08740 forbids the capability outright and no
+        // pipeline could be built from such a module. lavapipe reports it; Adreno and Mali both
+        // report VK_FALSE, so on every real mobile device this is false and the demotion runs
+        // exactly as it always has.
+        m_dynamicParameters.SupportsShaderFloat64 = m_vulkanCaps.SupportsShaderFloat64;
+        // shaderTessellationAndGeometryPointSize, both stage families from the one feature.
+        // False arms the shared phase-B point-size demotion, whose modules then carry no
+        // TessellationPointSize/GeometryPointSize capability and build without the feature.
+        // MOBILEGL_POINT_SIZE_DEMOTION=1 pretends it is absent so the demotion can be
+        // exercised on a healthy driver (lavapipe advertises the feature); =0 restores the
+        // detected answer's declines.
+        {
+            Bool supportsStagePointSize = m_vulkanCaps.SupportsTessellationAndGeometryPointSize;
+            switch (MG_Config::Features.PointSizeDemotion) {
+            case MG_Config::QuirkOverride::ForceOn:
+                MGLOG_I("DirectVulkan: MOBILEGL_POINT_SIZE_DEMOTION=1 - treating tessellation/geometry "
+                        "gl_PointSize as unhosted so the demotion runs on this driver");
+                supportsStagePointSize = false;
+                break;
+            case MG_Config::QuirkOverride::ForceOff:
+                MGLOG_I("DirectVulkan: MOBILEGL_POINT_SIZE_DEMOTION=0 - keeping the built-in and the "
+                        "plain declines regardless of the device feature");
+                supportsStagePointSize = true;
+                break;
+            case MG_Config::QuirkOverride::Auto:
+                break;
+            }
+            m_dynamicParameters.SupportsTessellationPointSize = supportsStagePointSize;
+            m_dynamicParameters.SupportsGeometryPointSize = supportsStagePointSize;
+        }
+        // Never, on any device, and DELIBERATELY NOT COUPLED to the line above even though it
+        // once tracked the same feature. It used to, because a `dvec` input needed Float64 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.
+        //
+        // Re-coupling it does not work, and the reason is worth recording because it is not
+        // obvious: this flag decides the VkFormat from the VAO ATTRIBUTE alone, and the attribute
+        // does not know what the shader declared. glVertexAttribFormat(GL_DOUBLE) against a plain
+        // `in vec4` is not only legal but the common case
+        // (KHR-GL43.vertex_attrib_binding.basic-input-case4 does exactly that, and case5 adds
+        // normalized=GL_TRUE), and advanced-bindingUpdate feeds a dvec3 the same way - GL defines
+        // all of them as "doubles in memory, converted to float". Turning the flag on turns the
+        // narrowing OFF for every one of them and the attributes come back unfetched.
+        //
+        // What keeps the two halves honest instead is a per-MODULE decision: a vertex module that
+        // declares a 64-bit float INPUT is demoted whole, even where the backend has native fp64,
+        // so `dvec` inputs are `vec` inputs on this backend exactly as they always were. See
+        // ShaderCompiler::SanitizeAndOptimizeBinary.
+        m_dynamicParameters.SupportsFloat64VertexAttributes = false;
         m_dynamicParameters.MaxShaderStorageBlockSize =
             std::min(m_vulkanCaps.MaxShaderStorageBlockSize, kMaxAdvertisedShaderStorageBlockSize);
         if (m_vulkanCaps.SupportsShaderSubgroup) {
             m_dynamicParameters.SubgroupSize = m_vulkanCaps.SubgroupSize;
             m_dynamicParameters.SubgroupSupportedStages = mapShaderStages(m_vulkanCaps.SubgroupSupportedStages);
-            m_dynamicParameters.SubgroupSupportedFeatures = mapSubgroupFeatures(m_vulkanCaps.SubgroupSupportedOperations);
+            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;
@@ -791,8 +1155,34 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
         if (m_dynamicParameters.MaxShaderStorageBlockSize != m_vulkanCaps.MaxShaderStorageBlockSize) {
             MGLOG_I("DirectVulkan: clamped GL_MAX_SHADER_STORAGE_BLOCK_SIZE from %zu to %zu",
-                    m_vulkanCaps.MaxShaderStorageBlockSize,
-                    m_dynamicParameters.MaxShaderStorageBlockSize);
+                    m_vulkanCaps.MaxShaderStorageBlockSize, m_dynamicParameters.MaxShaderStorageBlockSize);
+        }
+        switch (m_vulkanCaps.VendorId) {
+        case 0x5143u: // VK_VENDOR_ID: Qualcomm
+            m_dynamicParameters.GpuVendor = GpuVendorKind::Qualcomm;
+            break;
+        case 0x13B5u: // ARM
+            m_dynamicParameters.GpuVendor = GpuVendorKind::Arm;
+            break;
+        case 0x10DEu: // NVIDIA
+            m_dynamicParameters.GpuVendor = GpuVendorKind::Nvidia;
+            break;
+        case 0x1002u: // AMD
+            m_dynamicParameters.GpuVendor = GpuVendorKind::Amd;
+            break;
+        case 0x8086u: // Intel
+            m_dynamicParameters.GpuVendor = GpuVendorKind::Intel;
+            break;
+        case 0x1010u: // Imagination
+            m_dynamicParameters.GpuVendor = GpuVendorKind::ImgTec;
+            break;
+        case 0x10005u: // Mesa software (lavapipe)
+        case 0x1AE0u:  // Google (SwiftShader)
+            m_dynamicParameters.GpuVendor = GpuVendorKind::Software;
+            break;
+        default:
+            m_dynamicParameters.GpuVendor = GpuVendorKind::Unknown;
+            break;
         }
     }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/BackendObject_DirectVulkan.h

@@ -62,19 +62,21 @@ 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
-    // subgroup, no timer queries). A live backend copies this in its constructor and
+    // versions, ExtraVendor) with the baseline extension advertisement (no runtime-gated
+    // 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.
     const RendererInfo& GetRendererIdentity();
 
     // The full OpenGL extension list Magma advertises (glGetString(GL_EXTENSIONS)) for
-    // a device with the given raw capabilities. The MOBILEGL_DISABLE_SUBGROUP and
+    // a device with the given raw capabilities. The MOBILEGL_MAGMA_DISABLE_SUBGROUP and
     // 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,
+                                                  Bool cubeMapArraySupported);
 
     // 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.h

@@ -12,7 +12,7 @@
 #include "Renderer/VulkanRenderer.h"
 
 namespace MobileGL::MG_Backend::DirectVulkan {
-    extern UniquePtr<VulkanRenderer> pVulkanRenderer;
+    extern UniquePtr<VulkanRenderer>& pVulkanRenderer;
 
     // Generation of the live VulkanRenderer instance, mirroring DirectGLES's
     // g_syncContextGeneration. BackendObject_DirectVulkan bumps it wherever
@@ -23,12 +23,22 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     Uint64 GetRendererGeneration();
     void BumpRendererGeneration();
 
+    // Drops every cached program-resource reflection entry (CPU-side strings/vectors
+    // only, no Vulkan handles). Called at EGL teardown next to the renderer reset;
+    // safe because GL calls are serialized in this codebase, and any still-live
+    // program rebuilds its entry from the retained generated SPIR-V on demand.
+    void ClearProgramResourceCaches();
+
     void ClearBufferfi(GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
     void ClearBufferfv(GLenum buffer, GLint drawbuffer, const GLfloat* value);
     void ClearBufferuiv(GLenum buffer, GLint drawbuffer, const GLuint* value);
     void ClearBufferiv(GLenum buffer, GLint drawbuffer, const GLint* value);
     void ClearNamedFramebufferfv(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer, GLenum buffer,
                                  GLint drawbuffer, const GLfloat* value);
+    void ClearNamedFramebufferiv(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer, GLenum buffer,
+                                 GLint drawbuffer, const GLint* value);
+    void ClearNamedFramebufferuiv(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer, GLenum buffer,
+                                  GLint drawbuffer, const GLuint* value);
     void ClearNamedFramebufferfi(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer, GLenum buffer,
                                  GLint drawbuffer, GLfloat depth, GLint stencil);
     void Clear(GLbitfield mask);
@@ -72,9 +82,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                         GLsizei height, GLint border);
     void CopyTexSubImage2D(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width,
                            GLsizei height);
-    void CopyImageSubData(const SharedPtr<MG_State::GLState::ITextureObject>& srcTexture,
+    void CopyImageSubData(const CopyImageEndpoint& src,
                           GLenum srcTarget, GLint srcLevel, GLint srcX, GLint srcY, GLint srcZ,
-                          const SharedPtr<MG_State::GLState::ITextureObject>& dstTexture,
+                          const CopyImageEndpoint& dst,
                           GLenum dstTarget, GLint dstLevel, GLint dstX, GLint dstY, GLint dstZ,
                           GLsizei srcWidth, GLsizei srcHeight, GLsizei srcDepth);
     void GenerateMipmap(GLenum target);
@@ -87,15 +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);
-    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 ShaderStorageBlockBinding(GLuint program, const GLchar* storageBlockName, 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,
@@ -117,6 +119,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     // only while a live renderer exists whose device can actually time.
     Bool IsTimerQuerySupported();
     BackendQueryHandle BeginTimeElapsedQuery();
+    BackendQueryHandle BeginXfbPrimitivesQuery(Bool generated);
+    void EndXfbPrimitivesQuery(BackendQueryHandle query);
+    BackendQueryHandle BeginOcclusionQuery();
+    void EndOcclusionQuery(BackendQueryHandle query);
     void EndTimeElapsedQuery(BackendQueryHandle query);
     BackendQueryHandle QueryCounterTimestamp();
     Bool IsQueryResultAvailable(BackendQueryHandle query);

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

@@ -16,18 +16,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_device = device;
         m_commandPool = commandPool;
 
-        Vector<VkCommandBuffer> commandBuffers(frameCount, VK_NULL_HANDLE);
+        Vector<VkCommandBuffer> commandBuffers(frameCount * 2, VK_NULL_HANDLE);
         VkCommandBufferAllocateInfo allocInfo{};
         allocInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
         allocInfo.commandPool = commandPool;
         allocInfo.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
-        allocInfo.commandBufferCount = frameCount;
+        allocInfo.commandBufferCount = frameCount * 2;
         VkResult result = vkAllocateCommandBuffers(device, &allocInfo, commandBuffers.data());
         if (result != VK_SUCCESS) {
             return result;
         }
         for (Uint32 i = 0; i < frameCount; ++i) {
             m_frames[i].commandBuffer = commandBuffers[i];
+            m_frames[i].preCommandBuffer = commandBuffers[frameCount + i];
         }
 
         VkSemaphoreCreateInfo semaphoreInfo{VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO};
@@ -47,9 +48,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
     void FrameContext::Destroy(VkDevice device, VkCommandPool commandPool) {
         const Uint32 frameCount = static_cast<Uint32>(m_frames.size());
-        Vector<VkCommandBuffer> commandBuffers(frameCount, VK_NULL_HANDLE);
+        Vector<VkCommandBuffer> commandBuffers(frameCount * 2, VK_NULL_HANDLE);
         for (Uint32 i = 0; i < frameCount; ++i) {
             commandBuffers[i] = m_frames[i].commandBuffer;
+            commandBuffers[frameCount + i] = m_frames[i].preCommandBuffer;
         }
 
         for (Uint32 i = 0; i < frameCount; ++i) {
@@ -60,7 +62,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             for (auto& frame : m_frames) {
                 FreeRetiredCommandBuffers(frame);
             }
-            vkFreeCommandBuffers(device, commandPool, frameCount, commandBuffers.data());
+            vkFreeCommandBuffers(device, commandPool, frameCount * 2, commandBuffers.data());
         }
         m_frames.clear();
         currentFrameIndex = 0;
@@ -87,6 +89,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         currentFrameIndex = (currentFrameIndex + 1) % static_cast<Uint32>(m_frames.size());
         GetCurrent().isCommandRecording = false;
         GetCurrent().hasCommandBufferRecorded = false;
+        GetCurrent().isPreCommandRecording = false;
+        GetCurrent().hasPreCommandBufferRecorded = false;
     }
 
     VkCommandBuffer& FrameContext::BeginCommandRecording(VkCommandBufferUsageFlags flags,
@@ -118,6 +122,41 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         frame.hasCommandBufferRecorded = true;
     }
 
+    VkCommandBuffer FrameContext::BeginPreCommandRecording() {
+        auto& frame = GetCurrent();
+        if (frame.isPreCommandRecording) {
+            return frame.preCommandBuffer;
+        }
+        MOBILEGL_ASSERT(!frame.hasPreCommandBufferRecorded,
+                        "BeginPreCommandRecording: a recorded pre stream is still awaiting submission");
+        VK_VERIFY(vkResetCommandBuffer(frame.preCommandBuffer, 0), "BeginPreCommandRecording, vkResetCommandBuffer");
+        VkCommandBufferBeginInfo beginInfo{};
+        beginInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
+        VK_VERIFY(vkBeginCommandBuffer(frame.preCommandBuffer, &beginInfo),
+                  "BeginPreCommandRecording, vkBeginCommandBuffer");
+        frame.isPreCommandRecording = true;
+        return frame.preCommandBuffer;
+    }
+
+    void FrameContext::EndPreCommandRecordingIfOpen() {
+        auto& frame = GetCurrent();
+        if (!frame.isPreCommandRecording) {
+            return;
+        }
+        VK_VERIFY(vkEndCommandBuffer(frame.preCommandBuffer), "EndPreCommandRecordingIfOpen, vkEndCommandBuffer");
+        frame.isPreCommandRecording = false;
+        frame.hasPreCommandBufferRecorded = true;
+    }
+
+    void FrameContext::AbandonPreCommandRecording() {
+        auto& frame = GetCurrent();
+        if (frame.isPreCommandRecording) {
+            VK_VERIFY(vkEndCommandBuffer(frame.preCommandBuffer), "AbandonPreCommandRecording, vkEndCommandBuffer");
+        }
+        frame.isPreCommandRecording = false;
+        frame.hasPreCommandBufferRecorded = false;
+    }
+
     VkResult FrameContext::InitializeSwapchainSemaphores(VkDevice device, Uint32 swapchainImageCount) {
         DestroySwapchainSemaphores(device);
         if (swapchainImageCount == 0) {
@@ -150,12 +189,30 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
     Bool FrameContext::TransitionToPresent(VkImage image, VkImageLayout oldLayout, VkImageLayout presentLayout) {
         auto& frame = GetCurrent();
-        if (frame.hasCommandBufferRecorded || frame.isCommandRecording || oldLayout == presentLayout ||
-            oldLayout == VK_IMAGE_LAYOUT_SHARED_PRESENT_KHR) {
+        if (oldLayout == presentLayout || oldLayout == VK_IMAGE_LAYOUT_SHARED_PRESENT_KHR) {
             return false;
         }
 
-        auto& commandBuffer = BeginCommandRecording();
+        // The barrier belongs in the frame's own recording. Bailing out because
+        // something was already recorded (the previous behaviour) dropped the
+        // transition entirely for every frame that never ran a default-framebuffer
+        // render pass - the only other thing that carries the image to
+        // PRESENT_SRC_KHR, via that pass's finalLayout - so the swapchain image was
+        // handed to the WSI still in the layout it was acquired in.
+        // A closed-but-unsubmitted buffer can only come from a submit that already
+        // failed (SubmitPendingCommandBuffer leaves the flag set on error), and
+        // appending to it is illegal while reopening would reset the frame's own
+        // 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_ONCE("TransitionToPresent: command buffer already closed; skipping the present barrier");
+            return false;
+        }
+
+        // Reopening a recording here would vkResetCommandBuffer this frame's own
+        // commands away, so append to the open one and let the caller close it.
+        const Bool openedRecording = !frame.isCommandRecording;
+        VkCommandBuffer commandBuffer = openedRecording ? BeginCommandRecording() : frame.commandBuffer;
 
         VkImageMemoryBarrier presentBarrier{};
         presentBarrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
@@ -174,7 +231,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         vkCmdPipelineBarrier(commandBuffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, 0, 0,
                              nullptr, 0, nullptr, 1, &presentBarrier);
 
-        EndCommandRecording();
+        if (openedRecording) {
+            EndCommandRecording();
+        }
         return true;
     }
 
@@ -182,17 +241,27 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                                                Uint32 swapchainImageIndex) const {
         const auto& frame = GetCurrent();
         MOBILEGL_ASSERT(!frame.isCommandRecording, "GetSubmitInfo called while command buffer recording is still active");
+        MOBILEGL_ASSERT(!frame.isPreCommandRecording,
+                        "GetSubmitInfo called while the pre-pass stream is still recording");
         AssertValidSwapchainImageIndex(swapchainImageIndex);
         SubmitInfoPacket packet{};
         packet.waitSemaphore = frame.imageAvailableSemaphore;
         packet.signalSemaphore = m_swapchainImageRenderFinishedSemaphores[swapchainImageIndex];
-        packet.commandBuffer = frame.commandBuffer;
+
+        Uint32 commandBufferCount = 0;
+        // The pre-pass stream executes strictly before the frame's commands.
+        if (frame.hasPreCommandBufferRecorded) {
+            packet.commandBuffers[commandBufferCount++] = frame.preCommandBuffer;
+        }
+        if (shouldSubmitCommandBuffer) {
+            packet.commandBuffers[commandBufferCount++] = frame.commandBuffer;
+        }
 
         packet.submitInfo.waitSemaphoreCount = frame.imageAvailableSemaphoreConsumed ? 0U : 1U;
         packet.submitInfo.pWaitSemaphores = frame.imageAvailableSemaphoreConsumed ? nullptr : &packet.waitSemaphore;
         packet.submitInfo.pWaitDstStageMask = frame.imageAvailableSemaphoreConsumed ? nullptr : &packet.waitDstStageMask;
-        packet.submitInfo.commandBufferCount = shouldSubmitCommandBuffer ? 1U : 0U;
-        packet.submitInfo.pCommandBuffers = shouldSubmitCommandBuffer ? &packet.commandBuffer : nullptr;
+        packet.submitInfo.commandBufferCount = commandBufferCount;
+        packet.submitInfo.pCommandBuffers = commandBufferCount > 0 ? packet.commandBuffers : nullptr;
         packet.submitInfo.signalSemaphoreCount = 1;
         packet.submitInfo.pSignalSemaphores = &packet.signalSemaphore;
         return packet;
@@ -227,12 +296,21 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         result = vkAcquireNextImageKHR(device, swapchain, timeout, frame.imageAvailableSemaphore, acquireFence,
                                        &outImageIndex);
-        if (result != VK_SUCCESS) {
+        // VK_SUBOPTIMAL_KHR is a success code: an image *was* acquired and
+        // imageAvailableSemaphore *will* be signaled. Bailing out on it skipped both
+        // the consumed-flag reset (leaving a stale "already consumed", so the next
+        // submit never waited on the pending signal) and the fence reset (leaving
+        // the slot's fence signaled for the next submit to reuse). Only a genuine
+        // failure - VK_ERROR_OUT_OF_DATE_KHR and friends, where nothing is acquired
+        // and nothing is signaled - skips the bookkeeping.
+        if (result != VK_SUCCESS && result != VK_SUBOPTIMAL_KHR) {
             return result;
         }
 
         frame.imageAvailableSemaphoreConsumed = false;
-        return vkResetFences(device, 1, &frame.imageInFlightFence);
+        const VkResult resetResult = vkResetFences(device, 1, &frame.imageInFlightFence);
+        // Hand the acquire's own code back so the caller can schedule a rebuild.
+        return resetResult == VK_SUCCESS ? result : resetResult;
     }
 
     Uint32 FrameContext::GetCurrentFrameIndex() const {
@@ -247,12 +325,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_recordingObserver = observer;
     }
 
-    VkResult FrameContext::RetireCurrentCommandBuffer() {
+    VkResult FrameContext::RetireCurrentCommandBuffer(Bool retirePreCommandBuffer) {
         MOBILEGL_ASSERT(m_device != VK_NULL_HANDLE && m_commandPool != VK_NULL_HANDLE,
                         "RetireCurrentCommandBuffer requires an initialized FrameContext");
         auto& frame = GetCurrent();
         MOBILEGL_ASSERT(!frame.isCommandRecording,
                         "RetireCurrentCommandBuffer called while the command buffer is still recording");
+        MOBILEGL_ASSERT(!frame.isPreCommandRecording,
+                        "RetireCurrentCommandBuffer called while the pre-pass stream is still recording");
 
         VkCommandBufferAllocateInfo allocInfo{};
         allocInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
@@ -260,11 +340,23 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         allocInfo.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
         allocInfo.commandBufferCount = 1;
         VkCommandBuffer replacement = VK_NULL_HANDLE;
-        const VkResult result = vkAllocateCommandBuffers(m_device, &allocInfo, &replacement);
+        VkResult result = vkAllocateCommandBuffers(m_device, &allocInfo, &replacement);
         if (result != VK_SUCCESS) {
             return result;
         }
-        frame.retiredCommandBuffers.push_back(frame.commandBuffer);
+        if (retirePreCommandBuffer) {
+            VkCommandBuffer preReplacement = VK_NULL_HANDLE;
+            result = vkAllocateCommandBuffers(m_device, &allocInfo, &preReplacement);
+            if (result != VK_SUCCESS) {
+                vkFreeCommandBuffers(m_device, m_commandPool, 1, &replacement);
+                return result;
+            }
+            frame.retiredCommandBuffers.push_back({frame.preCommandBuffer, frame.lastSubmitIndex});
+            frame.preCommandBuffer = preReplacement;
+        }
+        // lastSubmitIndex was just written by the renderer for the submission
+        // that carried this command buffer.
+        frame.retiredCommandBuffers.push_back({frame.commandBuffer, frame.lastSubmitIndex});
         frame.commandBuffer = replacement;
         return VK_SUCCESS;
     }
@@ -274,12 +366,40 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return;
         }
         if (m_device != VK_NULL_HANDLE && m_commandPool != VK_NULL_HANDLE) {
-            vkFreeCommandBuffers(m_device, m_commandPool, static_cast<Uint32>(frame.retiredCommandBuffers.size()),
-                                 frame.retiredCommandBuffers.data());
+            for (const auto& retired : frame.retiredCommandBuffers) {
+                vkFreeCommandBuffers(m_device, m_commandPool, 1, &retired.commandBuffer);
+            }
         }
         frame.retiredCommandBuffers.clear();
     }
 
+    void FrameContext::FreeRetiredCommandBuffersCompletedUpTo(Uint64 completedSubmitIndex) {
+        if (m_device == VK_NULL_HANDLE || m_commandPool == VK_NULL_HANDLE) {
+            return;
+        }
+        for (auto& frame : m_frames) {
+            // Retired buffers are appended in submit order, so the completed
+            // ones form a prefix.
+            SizeT completedCount = 0;
+            while (completedCount < frame.retiredCommandBuffers.size() &&
+                   frame.retiredCommandBuffers[completedCount].submitIndex <= completedSubmitIndex) {
+                vkFreeCommandBuffers(m_device, m_commandPool, 1,
+                                     &frame.retiredCommandBuffers[completedCount].commandBuffer);
+                ++completedCount;
+            }
+            if (completedCount > 0) {
+                frame.retiredCommandBuffers.erase(frame.retiredCommandBuffers.begin(),
+                                                  frame.retiredCommandBuffers.begin() + completedCount);
+            }
+        }
+    }
+
+    void FrameContext::FreeAllRetiredCommandBuffers() {
+        for (auto& frame : m_frames) {
+            FreeRetiredCommandBuffers(frame);
+        }
+    }
+
     void FrameContext::AssertValidFrameIndex(Uint32 frameIndex) const {
         MOBILEGL_ASSERT(frameIndex < m_frames.size(), "FrameContext index out of range");
     }

MobileGL/MG_Backend/DirectVulkan/Renderer/FrameContext.h

@@ -29,7 +29,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VkPipelineStageFlags waitDstStageMask = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
             VkSemaphore waitSemaphore = VK_NULL_HANDLE;
             VkSemaphore signalSemaphore = VK_NULL_HANDLE;
-            VkCommandBuffer commandBuffer = VK_NULL_HANDLE;
+            // [0] = pre-pass command buffer (when recorded), then the frame
+            // command buffer; submitInfo.pCommandBuffers points here.
+            VkCommandBuffer commandBuffers[2] = {VK_NULL_HANDLE, VK_NULL_HANDLE};
             VkSubmitInfo submitInfo{VK_STRUCTURE_TYPE_SUBMIT_INFO};
         };
 
@@ -40,17 +42,35 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VkPresentInfoKHR presentInfo{VK_STRUCTURE_TYPE_PRESENT_INFO_KHR};
         };
 
+        // A command buffer submitted mid-frame (FlushPendingCommands), tagged
+        // with the submit-tracker index it was submitted under so it can be
+        // freed as soon as that submission is observed complete - without
+        // waiting for the slot's fence to be waited again (present-less flush
+        // loops never wait it).
+        struct RetiredCommandBuffer {
+            VkCommandBuffer commandBuffer = VK_NULL_HANDLE;
+            Uint64 submitIndex = 0;
+        };
+
         struct FrameData {
             VkCommandBuffer commandBuffer = VK_NULL_HANDLE;
+            // Pre-pass work stream: out-of-pass commands (deferred clear
+            // materialization, sampled-layout transitions) for resources the
+            // frame's recording has not touched yet. Submitted immediately
+            // BEFORE commandBuffer in the same vkQueueSubmit, so recording
+            // into it never has to split the frame's active render pass.
+            VkCommandBuffer preCommandBuffer = VK_NULL_HANDLE;
             VkSemaphore imageAvailableSemaphore = VK_NULL_HANDLE;
             VkFence imageInFlightFence = VK_NULL_HANDLE;
             Bool isCommandRecording = false;
             Bool hasCommandBufferRecorded = false;
+            Bool isPreCommandRecording = false;
+            Bool hasPreCommandBufferRecorded = false;
             Bool imageAvailableSemaphoreConsumed = false;
-            // Command buffers submitted mid-frame (FlushPendingCommands) whose
-            // execution is only known complete once this slot's fence has been
-            // waited again; freed at that point.
-            Vector<VkCommandBuffer> retiredCommandBuffers;
+            // Command buffers submitted mid-frame (FlushPendingCommands),
+            // appended in submit order; freed once their submission is known
+            // complete (fence wait or completion poll).
+            Vector<RetiredCommandBuffer> retiredCommandBuffers;
             // Submit-tracker index of this slot's most recent queue submission
             // (written by the renderer at submit time).
             Uint64 lastSubmitIndex = 0;
@@ -67,6 +87,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkCommandBuffer& BeginCommandRecording(VkCommandBufferUsageFlags flags = 0,
                                                const VkCommandBufferInheritanceInfo* pInheritanceInfo = nullptr);
         void EndCommandRecording();
+        // Lazily opens the pre-pass work stream (see FrameData::preCommandBuffer).
+        VkCommandBuffer BeginPreCommandRecording();
+        // Closes the pre stream if open, marking it for submission ahead of the
+        // frame command buffer. Safe to call when it never opened.
+        void EndPreCommandRecordingIfOpen();
+        // Drops an in-progress or recorded-but-unsubmitted pre stream (dropped
+        // frame recordings, swapchain recreation).
+        void AbandonPreCommandRecording();
         VkResult InitializeSwapchainSemaphores(VkDevice device, Uint32 swapchainImageCount);
         void DestroySwapchainSemaphores(VkDevice device);
         Bool TransitionToPresent(VkImage image, VkImageLayout oldLayout,
@@ -79,8 +107,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // Parks the current (already ended and submitted) command buffer on the
         // slot's retired list and installs a freshly allocated one, so recording
         // can restart while the submitted buffer is still executing. Retired
-        // buffers are freed after the slot's fence is next waited.
-        VkResult RetireCurrentCommandBuffer();
+        // buffers are freed after the slot's fence is next waited, or as soon
+        // as their submission is observed complete.
+        VkResult RetireCurrentCommandBuffer(Bool retirePreCommandBuffer = false);
+
+        // Frees every retired command buffer whose tagged submission index is
+        // known complete. Driven by the renderer's submit tracker on completion
+        // events (fence waits and non-blocking polls), so present-less flush
+        // loops reclaim their buffers without any extra wait.
+        void FreeRetiredCommandBuffersCompletedUpTo(Uint64 completedSubmitIndex);
+        // Frees every slot's retired command buffers. Only valid when the
+        // caller has proven every queue submission complete.
+        void FreeAllRetiredCommandBuffers();
 
         Uint32 GetCurrentFrameIndex() const;
         Uint32 GetFrameCount() const;

MobileGL/MG_Backend/DirectVulkan/Renderer/PipelineFactory.cpp

@@ -8,6 +8,8 @@
 
 #include "PipelineFactory.h"
 
+#include <algorithm>
+
 namespace MobileGL::MG_Backend::DirectVulkan {
     static const char* PrimitiveTopologyToString(VkPrimitiveTopology topology) {
         switch (topology) {
@@ -108,6 +110,81 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                   "vkCreatePipelineCache");
     }
 
+    // Must be called once, before any pipeline is created: the flag is not part of the
+    // pipeline hash, so flipping it mid-life would serve cached pipelines built under the
+    // old value.
+    void PipelineFactory::SetSuppressBlendedDepthWrite(Bool enabled) {
+        s_suppressBlendedDepthWrite = enabled;
+    }
+
+    Bool PipelineFactory::ShouldSuppressBlendedDepthWriteForDevice(MG_Config::QuirkOverride quirkOverride,
+                                                                   Uint32 vendorId) {
+        static constexpr Uint32 kVendorIdQualcomm = 0x5143;
+        switch (quirkOverride) {
+        case MG_Config::QuirkOverride::ForceOn:
+            return true;
+        case MG_Config::QuirkOverride::ForceOff:
+            return false;
+        case MG_Config::QuirkOverride::Auto:
+        default:
+            return vendorId == kVendorIdQualcomm;
+        }
+    }
+
+    namespace {
+        // MIN/MAX extremum blending: the signature of a depth-bounds accumulation pass
+        // (MC 26.3 OIT writes vec4(-linD, linD, deviceZ, 0) under GL_MAX while writing
+        // depth for its equality chain). MIN/MAX ignore blend factors per the Vulkan spec.
+        //
+        // Deliberately the ONLY shape stripped. A quirk should touch as little unrelated
+        // content as possible, and a trace sweep of every fixture showed the wider
+        // alternatives all cost more than they fix:
+        //   - additive ONE+ONE with a depth write matched zero draws of the 26.3 chain
+        //     (its transmittance/accumulate passes disable depth writes themselves) - the
+        //     only real content it caught was harmless additive glow effects (Create);
+        //   - sorted-transparency "over" blends (SRC_ALPHA-style) are order-dependent,
+        //     drawn once per surface, and rely on their depth writes for occlusion;
+        //   - separate-alpha accumulation over an over-blending color channel has no
+        //     known pairing with a depth-equality chain (color channel only, see tests).
+        // If a future workload pairs another blend shape with an equality chain, widen
+        // this with that evidence in hand rather than pre-emptively.
+        Bool IsAccumulationBlend(const VkPipelineColorBlendAttachmentState& attachment) {
+            return attachment.colorBlendOp == VK_BLEND_OP_MIN ||
+                   attachment.colorBlendOp == VK_BLEND_OP_MAX;
+        }
+    } // namespace
+
+    Bool PipelineFactory::ShouldSuppressDepthWrite(const PipelineCreatePayload& payload) {
+        if (!payload.depthWriteEnable) {
+            return false;
+        }
+        // A shader that assigns gl_FragDepth supplies depth itself rather than taking the
+        // pipeline's interpolated Z, so a driver that varies the vertex position math
+        // between pipelines cannot desynchronize it. (A gl_FragDepth = gl_FragCoord.z
+        // passthrough is the exception that stays exposed; no known content pairs one with
+        // an equality chain, and 26.3's composite is a genuine computed-depth writer.)
+        if (payload.fragmentReplacesDepth) {
+            return false;
+        }
+        for (Uint32 i = 0; i < payload.colorAttachmentCount; ++i) {
+            const VkPipelineColorBlendAttachmentState& attachment = payload.colorBlendAttachments[i];
+            if (attachment.blendEnable != VK_TRUE) {
+                continue;
+            }
+            // All color writes masked: blending is moot (depth-prepass pattern that left
+            // GL_BLEND enabled); stripping the depth write would delete the whole prepass.
+            if (attachment.colorWriteMask == 0) {
+                continue;
+            }
+            // Any attachment qualifies, not just attachment 0: the 26.3 transmittance pass
+            // accumulates into a 2-target MRT and must stay stripped.
+            if (IsAccumulationBlend(attachment)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
     PipelineFactory::~PipelineFactory() {
         DestroyAll();
         if (m_pipelineCache != VK_NULL_HANDLE) {
@@ -124,13 +201,22 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.renderPass, sizeof(payload.renderPass)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.colorAttachmentCount, sizeof(payload.colorAttachmentCount)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.rasterizationSamples, sizeof(payload.rasterizationSamples)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.sampleShadingEnable, sizeof(payload.sampleShadingEnable)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.minSampleShading, sizeof(payload.minSampleShading)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.sampleMask, sizeof(payload.sampleMask)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.subpass, sizeof(payload.subpass)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.topology, sizeof(payload.topology)));
         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.passthroughTessControlKey,
+                                   sizeof(payload.passthroughTessControlKey)));
+        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)));
@@ -152,6 +238,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             XXH64_update(m_hashState, &payload.backStencilDepthFailOp, sizeof(payload.backStencilDepthFailOp)));
         XXHASH_VERIFY(
             XXH64_update(m_hashState, &payload.backStencilCompareOp, sizeof(payload.backStencilCompareOp)));
+        XXHASH_VERIFY(
+            XXH64_update(m_hashState, &payload.fragmentReplacesDepth, sizeof(payload.fragmentReplacesDepth)));
         if (payload.colorAttachmentCount > 0) {
             XXHASH_VERIFY(XXH64_update(
                 m_hashState,
@@ -165,23 +253,125 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         const HashType hash = ComputeHash(payload);
         auto it = m_cache.find(hash);
         if (it != m_cache.end()) {
-            return it->second;
+            it->second.lastUsedFrame = m_frameCounter;
+            return it->second.pipeline;
         }
 
         VkPipeline pipeline = CreatePipeline(payload);
-        m_cache.emplace(hash, pipeline);
+        // 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;
     }
 
     void PipelineFactory::DestroyAll() {
         for (auto& pair : m_cache) {
-            if (pair.second != VK_NULL_HANDLE) {
-                vkDestroyPipeline(m_device, pair.second, nullptr);
+            if (pair.second.pipeline != VK_NULL_HANDLE) {
+                vkDestroyPipeline(m_device, pair.second.pipeline, nullptr);
             }
         }
         m_cache.clear();
     }
 
+    Uint32 PipelineFactory::OnFrameBoundary() {
+        ++m_frameCounter;
+
+        // Sweep cadence and retire age mirror VkRenderPassManager::OnPresent: an entry
+        // idle for more than kRetireAgeFrames frame boundaries cannot be referenced by
+        // any in-flight command buffer (frames-in-flight <= MOBILEGL_MAGMA_FRAMESINFLIGHT),
+        // so immediate vkDestroyPipeline is safe. The caller must drop its "last
+        // pipeline" memo when this returns non-zero: the memo can return a cached
+        // handle without touching this cache, so an evicted pipeline may still be
+        // memoized (present-less flush loops never reset the memo per frame).
+        constexpr Uint64 kSweepInterval = 256;
+        constexpr Uint64 kRetireAgeFrames = 1024;
+        if ((m_frameCounter % kSweepInterval) != 0) {
+            return 0;
+        }
+
+        Uint32 evicted = 0;
+        for (auto it = m_cache.begin(); it != m_cache.end();) {
+            if (m_frameCounter - it->second.lastUsedFrame > kRetireAgeFrames) {
+                if (it->second.pipeline != VK_NULL_HANDLE) {
+                    vkDestroyPipeline(m_device, it->second.pipeline, nullptr);
+                }
+                it = m_cache.erase(it);
+                ++evicted;
+            } else {
+                ++it;
+            }
+        }
+        if (evicted > 0) {
+            MGLOG_D("PipelineFactory::OnFrameBoundary: evicted %u idle pipelines (%zu remain)", evicted,
+                    m_cache.size());
+        }
+        return evicted;
+    }
+
+    Uint32 PipelineFactory::EvictByRenderPasses(const Vector<VkRenderPass>& renderPasses) {
+        if (renderPasses.empty() || m_cache.empty()) {
+            return 0;
+        }
+        // Sorted-batch membership test keeps a mass eviction (shader-pack switch,
+        // dimension exit) at one O(cache * log batch) scan instead of one full scan
+        // per dying pass.
+        Vector<VkRenderPass> sortedPasses = renderPasses;
+        std::sort(sortedPasses.begin(), sortedPasses.end());
+        Uint32 evicted = 0;
+        for (auto it = m_cache.begin(); it != m_cache.end();) {
+            if (std::binary_search(sortedPasses.begin(), sortedPasses.end(), it->second.renderPass)) {
+                if (it->second.pipeline != VK_NULL_HANDLE) {
+                    vkDestroyPipeline(m_device, it->second.pipeline, nullptr);
+                }
+                it = m_cache.erase(it);
+                ++evicted;
+            } else {
+                ++it;
+            }
+        }
+        if (evicted > 0) {
+            MGLOG_D("PipelineFactory::EvictByRenderPasses: evicted %u pipelines for %zu destroyed render passes",
+                    evicted, sortedPasses.size());
+        }
+        return evicted;
+    }
+
+    Uint32 PipelineFactory::EvictByProgramHash(HashType programHash) {
+        Uint32 evicted = 0;
+        for (auto it = m_cache.begin(); it != m_cache.end();) {
+            if (it->second.programHash == programHash) {
+                if (it->second.pipeline != VK_NULL_HANDLE) {
+                    vkDestroyPipeline(m_device, it->second.pipeline, nullptr);
+                }
+                it = m_cache.erase(it);
+                ++evicted;
+            } else {
+                ++it;
+            }
+        }
+        if (evicted > 0) {
+            MGLOG_D("PipelineFactory::EvictByProgramHash: evicted %u pipelines for program hash 0x%llx",
+                    evicted, static_cast<unsigned long long>(programHash));
+        }
+        return evicted;
+    }
+
     VkPipeline PipelineFactory::CreatePipeline(const PipelineCreatePayload& payload) const {
         MOBILEGL_ASSERT(payload.stages != nullptr && !payload.stages->empty(), "PipelineFactory: stages are empty");
         MOBILEGL_ASSERT(payload.vertexInputState != nullptr, "PipelineFactory: vertexInputState is null");
@@ -216,9 +406,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         ia.topology = payload.topology;
         ia.primitiveRestartEnable = payload.primitiveRestartEnable ? VK_TRUE : VK_FALSE;
 
+        // Only a patch topology has a tessellation stage to configure; leaving the pointer null
+        // otherwise is what the spec expects.
+        VkPipelineTessellationStateCreateInfo tessellation{VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO};
+        tessellation.patchControlPoints = payload.patchControlPoints;
+
         VkPipelineViewportStateCreateInfo vpci{VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO};
-        vpci.viewportCount = 1;
-        vpci.scissorCount = 1;
+        // Both counts move together: GL has one scissor rectangle per viewport, and Vulkan
+        // 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;
@@ -227,9 +426,28 @@ 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;
+        ms.sampleShadingEnable = payload.sampleShadingEnable ? VK_TRUE : VK_FALSE;
+        // Ignored by Vulkan unless sampleShadingEnable is set, but written unconditionally so the
+        // struct's bytes match the hash the payload was keyed by.
+        ms.minSampleShading = payload.minSampleShading;
+        // GL_SAMPLE_MASK / glSampleMaski. Left at nullptr - which Vulkan reads as all-ones - until
+        // now, so glSampleMaski was a silent no-op on this backend while DirectGLES forwarded it.
+        // The pointer has to outlive the vkCreateGraphicsPipelines call, which the payload does.
+        ms.pSampleMask = payload.sampleMask;
 
         VkPipelineDepthStencilStateCreateInfo depthStencil{VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO};
         depthStencil.depthTestEnable = payload.depthTestEnable ? VK_TRUE : VK_FALSE;
@@ -258,17 +476,77 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         for (Uint32 i = 0; i < payload.colorAttachmentCount; ++i) {
             colorAttachments[i] = payload.colorBlendAttachments[i];
         }
+        // Suppress depth writes on accumulation-blended pipelines when the active driver
+        // cannot keep vertex positions invariant across the pipelines of a multi-pass
+        // depth-equality chain (see SetSuppressBlendedDepthWrite). The decision is narrowed
+        // in ShouldSuppressDepthWrite: sorted-transparency "over" blends (vanilla MC water),
+        // gl_FragDepth writers, and masked-out attachments keep their depth writes.
+        // This bakes the decision into the pipeline, which only works because depth write is
+        // static state here - adding VK_DYNAMIC_STATE_DEPTH_WRITE_ENABLE to kDynamicStates
+        // would let the record-time value override it and silently disable the quirk.
+        if (s_suppressBlendedDepthWrite && ShouldSuppressDepthWrite(payload)) {
+            depthStencil.depthWriteEnable = VK_FALSE;
+        }
         VkPipelineColorBlendStateCreateInfo blend{VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO};
         blend.logicOpEnable = payload.logicOpEnable ? VK_TRUE : VK_FALSE;
         blend.logicOp = payload.logicOp;
         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.pStages = payload.stages->data();
+        gpi.stageCount = static_cast<Uint32>(effectiveStages->size());
+        gpi.pStages = effectiveStages->data();
         gpi.pVertexInputState = payload.vertexInputState;
         gpi.pInputAssemblyState = &ia;
+        gpi.pTessellationState =
+            payload.topology == VK_PRIMITIVE_TOPOLOGY_PATCH_LIST ? &tessellation : nullptr;
         gpi.pViewportState = &vpci;
         gpi.pRasterizationState = &raster;
         gpi.pMultisampleState = &ms;
@@ -281,6 +559,13 @@ 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),
@@ -311,6 +596,36 @@ 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;
@@ -27,12 +37,55 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VkRenderPass renderPass = VK_NULL_HANDLE;
             Uint32 colorAttachmentCount = 1;
             VkSampleCountFlagBits rasterizationSamples = VK_SAMPLE_COUNT_1_BIT;
+            // glEnable(GL_SAMPLE_SHADING) + glMinSampleShading, which Vulkan bakes into the
+            // pipeline rather than exposing as dynamic state - so both are part of the pipeline's
+            // identity and both are hashed. The renderer leaves the enable false unless the
+            // device's sampleRateShading feature was enabled
+            // (VUID-VkPipelineMultisampleStateCreateInfo-sampleShadingEnable-00784).
+            Bool sampleShadingEnable = false;
+            Float minSampleShading = 0.0f;
+            // glEnable(GL_SAMPLE_MASK) + glSampleMaski, the fixed-function coverage mask, already
+            // reduced to what GL says this draw gets (VulkanRenderer::ResolveEffectiveSampleMask:
+            // all-ones unless the target is genuinely multisampled). Pipeline state like the two
+            // above - Vulkan has no dynamic sample mask before VK_EXT_extended_dynamic_state3 -
+            // so it is hashed with them, and all-ones has to keep producing the pipeline a null
+            // pSampleMask always did.
+            //
+            // TWO words, though GL only ever fills the first. GL_MAX_SAMPLE_MASK_WORDS is clamped
+            // to 1 on both backends, so glSampleMaski writes index 0 and nothing else - but the
+            // count Vulkan READS is ceil(rasterizationSamples / 32), which is 2 on a 64-sample
+            // target, and GetAdvertisedMaxSamples does not cap the driver's sample count. A
+            // single Uint32 here let such a pipeline read one word past the member (the next
+            // struct field). The second word is all-ones: full coverage for samples 32..63, which
+            // is the only honest answer when GL has no state describing them.
+            Uint32 sampleMask[2] = {0xffffffffu, 0xffffffffu};
             Uint32 subpass = 0;
             VkPrimitiveTopology topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST;
             Bool primitiveRestartEnable = false;
+            // GL_PATCH_VERTICES; only read for a PATCH_LIST topology.
+            Uint32 patchControlPoints = 3;
+            // ProgramFactory::ComputePassthroughTessControlKey of the synthesized pass-through
+            // tessellation control stage below, or 0 when this pipeline has none. Hashed, because
+            // the levels glPatchParameterfv set are compiled INTO that module and are not a
+            // function of the program or of patchControlPoints - see the note on
+            // passthroughTessControlStage.
+            Uint64 passthroughTessControlKey = 0;
+            // 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;
@@ -49,9 +102,26 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VkStencilOp backStencilPassOp = VK_STENCIL_OP_KEEP;
             VkStencilOp backStencilDepthFailOp = VK_STENCIL_OP_KEEP;
             VkCompareOp backStencilCompareOp = VK_COMPARE_OP_ALWAYS;
+            // The fragment module writes gl_FragDepth (SPIR-V DepthReplacing); exempts the
+            // pipeline from the blended depth-write quirk (see ShouldSuppressDepthWrite).
+            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 directly: it is a pure function of the program, of patchControlPoints and
+            // of the default tessellation levels - the first two of which ComputeHash already
+            // mixes in, and the third of which arrives through passthroughTessControlKey above.
+            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);
@@ -62,13 +132,69 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkPipeline GetOrCreatePipeline(const PipelineCreatePayload& payload);
         void DestroyAll();
 
+        // Frame boundary hook: ages the pipeline cache and destroys long-unused entries
+        // (their command buffers retired many frames ago), mirroring
+        // VkRenderPassManager::OnPresent's sweep. Returns the number of pipelines
+        // destroyed so the caller can drop any memoized VkPipeline handle.
+        Uint32 OnFrameBoundary();
+        // Destroys every cached pipeline hashed on one of `renderPasses`. Only safe
+        // when the caller guarantees GPU idleness for them - the render-pass manager
+        // calls this (via the renderer) for passes its own >1024-boundary-idle sweep
+        // just evicted, and a pipeline hashed on those handles is only ever bound by
+        // draws that also hit the render-pass entries. Also closes the handle-recycling
+        // hazard: a recycled VkRenderPass value must never serve a stale pipeline.
+        // Batched: one cache scan regardless of how many passes died in the sweep.
+        // Returns the number destroyed (callers invalidate memos when non-zero).
+        Uint32 EvictByRenderPasses(const Vector<VkRenderPass>& renderPasses);
+        // Destroys every cached pipeline built from the program with content hash
+        // `programHash`. Called from the ProgramFactory eviction path, which proves the
+        // same >1024-boundary idleness (the program's pipelines are only bound by draws
+        // that stamp its factory entry). Returns the number destroyed.
+        Uint32 EvictByProgramHash(HashType programHash);
+
+        // Driver quirk: suppress depth writes on accumulation-blended pipelines. Multi-pass
+        // depth-equality rendering (a blended prepass writes depth that later passes re-test
+        // with an equality-inclusive compare on the re-rasterized geometry) requires
+        // cross-pipeline position invariance that some mobile compilers do not provide, even
+        // with the SPIR-V Invariant decoration; whole primitives then drop out of the later
+        // passes. Only MIN/MAX extremum blends are stripped - the signature of such a
+        // chain's depth-bounds pass (MC 26.3 OIT), and per a fixture-wide trace sweep the
+        // only depth-writing shape the chain actually uses - so every other blend
+        // (sorted-transparency "over" like vanilla MC water, additive glows, ...) keeps
+        // its depth writes. Set at renderer initialization based on the active driver.
+        static void SetSuppressBlendedDepthWrite(Bool enabled);
+        static Bool IsSuppressBlendedDepthWriteEnabled() { return s_suppressBlendedDepthWrite; }
+        // Device gate for the quirk: ForceOn/ForceOff bypass detection, Auto enables it on
+        // the known-affected vendor (Qualcomm).
+        static Bool ShouldSuppressBlendedDepthWriteForDevice(MG_Config::QuirkOverride quirkOverride,
+                                                             Uint32 vendorId);
+        // Pure per-pipeline strip decision (exempts gl_FragDepth writers, masked-out and
+        // non-accumulation blends); combined with the device flag in CreatePipeline. Static
+        // and payload-only so tests can pin the contract without a VkDevice.
+        static Bool ShouldSuppressDepthWrite(const PipelineCreatePayload& payload);
+
     private:
+        struct PipelineCacheEntry {
+            VkPipeline pipeline = VK_NULL_HANDLE;
+            // The hashed inputs the eviction paths key on: programHash ties the entry to
+            // its ProgramFactory entry, renderPass records the exact handle the hash
+            // folded in (the hash is one-way, so targeted eviction needs them verbatim).
+            HashType programHash = 0;
+            VkRenderPass renderPass = VK_NULL_HANDLE;
+            // Frame-boundary counter value of the last GetOrCreatePipeline hit; drives
+            // cache eviction (see OnFrameBoundary).
+            Uint64 lastUsedFrame = 0;
+        };
+
         VkPipeline CreatePipeline(const PipelineCreatePayload& payload) const;
 
         VkDevice m_device = VK_NULL_HANDLE;
         const VulkanRendererConfig& m_config;
         VkPipelineCache m_pipelineCache = VK_NULL_HANDLE;
-        UnorderedMap<HashType, VkPipeline> m_cache;
+        UnorderedMap<HashType, PipelineCacheEntry> m_cache;
+        // Monotonic frame-boundary counter (bumped in OnFrameBoundary) for cache aging.
+        Uint64 m_frameCounter = 0;
         static inline XXH64_state_t* m_hashState = XXH64_createState();
+        static inline Bool s_suppressBlendedDepthWrite = false;
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

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 {
@@ -42,31 +49,129 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             SurfaceRotate90 = 1 << 2,
             SurfaceRotate180 = 1 << 3,
             SurfaceRotate270 = 1 << 4,
+            // Rewrites the fragment stage's implicit-LOD image samples to explicit LOD 0.
+            // Only ever set for a draw whose every sampler binding is clamped to a single mip
+            // level, which makes the two forms produce identical texels (the implicit lambda is
+            // clamped into [minLod, maxLod] = [0, 0] regardless of derivatives or bias).
+            ExplicitLod0Sampling = 1 << 5,
+            // Decorates the last vertex-processing stage's captured varyings with
+            // XfbBuffer/XfbStride/Offset (VK_EXT_transform_feedback). Set only for draws
+            // 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;
 
+        // The gl_PerVertex members a pass-through tessellation control stage may have to carry,
+        // in the order glslang declares them - which is the order a redeclaration must use.
+        // Which of them exist is a function of the neighbouring stage's GLSL VERSION
+        // (gl_CullDistance joins the block at #version 450), so the mask is read off that
+        // stage's SPIR-V rather than assumed. See ReflectPerVertexInputMembers.
+        enum class PerVertexMemberBit : Uint32 {
+            Position = 1u << 0,
+            PointSize = 1u << 1,
+            ClipDistance = 1u << 2,
+            CullDistance = 1u << 3,
+        };
+        // What a program parsed below #version 450 carries, and the fallback when a module's
+        // block cannot be read.
+        static constexpr Uint32 kDefaultPerVertexMembers =
+            static_cast<Uint32>(PerVertexMemberBit::Position) | static_cast<Uint32>(PerVertexMemberBit::PointSize) |
+            static_cast<Uint32>(PerVertexMemberBit::ClipDistance);
+
+        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 a descriptor ARRAY - a UBO or storage
+            // 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.
+            UnorderedMap<Uint32, Vector<Int>> arrayedUniformBlockIndicesByBinding;
             Vector<String> samplerNameByBinding;
             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;
             Vector<Int> storageBlockIndexByBinding;
+            // 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{};
@@ -75,6 +180,76 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             ShaderStage rasterizationProducerStage = ShaderStage::Unknown;
             Uint32 producerOutputComponentCount = 0;
             Uint32 fragmentInputComponentCount = 0;
+            // The fragment module declares the DepthReplacing execution mode (writes
+            // 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.
+            // True when this program was built AS a transform-feedback capture variant but its
+            // last pre-rasterization module does NOT carry the Xfb execution mode - so the
+            // renderer must decline the capture span instead of issuing
+            // vkCmdBeginTransformFeedbackEXT against it
+            // (VUID-vkCmdBeginTransformFeedbackEXT-None-04128).
+            //
+            // Two ways to get here, and neither is visible from GL state, which is all
+            // BeginXfbCaptureForDraw otherwise consults: the clip/XFB validation backstop had to
+            // rewind past the capture decoration, or XfbCaptureDecoratePass resolved none of the
+            // requested varyings and returned without changing anything (its own MGLOG_E path)
+            // while its runner still reported success. Both used to ship a non-Xfb module under
+            // an Xfb-flagged cache entry - the flag and the layout are part of the program cache
+            // key, so it was sticky for every later captured draw of the program, not a glitch.
+            Bool xfbCaptureDeclined = false;
+            // The program has a tessellation or geometry module declaring TessellationPointSize /
+            // GeometryPointSize on a device whose shaderTessellationAndGeometryPointSize feature
+            // is off, so a pipeline built from it is invalid usage
+            // (VUID-RuntimeSpirv-PointSize-06439). Its draws are refused in SetupDraw rather than
+            // handed to the driver - the same contract PipelineFactory's half-tessellated refusal
+            // implements one level up, and the counterpart of the DirectGLES arm that reports a
+            // driver with neither point-size extension by name.
+            //
+            // Sticky by construction, which is what makes ONE log line honest: the flag lives on
+            // the cache entry, so every later draw of the same program variant reads the same
+            // answer instead of re-deciding it.
+            Bool pointSizeCapabilityUnsupported = false;
+            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;
+            // Which gl_PerVertex members the evaluation stage's `in gl_PerVertex gl_in[]` block
+            // actually carries, as a PerVertexMemberBit mask read off its SPIR-V. The synthesized
+            // control stage has to redeclare the SAME shape: glslang appends gl_CullDistance to
+            // that block from #version 450 upward, so a 450/460 program - and every ESSL program,
+            // which the source processor rewrites to "#version 460 core" - carries four members
+            // where a 430 program carries three. A fixed three-member pass-through fed the
+            // evaluation stage a differently-shaped block, which is the black-frame-no-error case
+            // this whole family is written around.
+            Uint32 passthroughPerVertexMembers = 0;
+            // Frame-boundary counter value of the last GetOrCreateProgram hit; drives
+            // cache eviction (see OnFrameBoundary). Mutable: the draw snapshot's memoised
+            // entry pointer re-stamps use through a const reference (StampProgramUse).
+            mutable Uint64 lastUsedFrame = 0;
 
             static inline VkDevice s_device = VK_NULL_HANDLE;
 
@@ -85,11 +260,23 @@ 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);
+                arrayedUniformBlockIndicesByBinding = std::move(other.arrayedUniformBlockIndicesByBinding);
                 samplerNameByBinding = std::move(other.samplerNameByBinding);
                 samplerUniformLocationByBinding = std::move(other.samplerUniformLocationByBinding);
                 samplerTextureTargetByBinding = std::move(other.samplerTextureTargetByBinding);
@@ -99,6 +286,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                     std::move(other.storageImageUsesBindingFormatByBinding);
                 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;
@@ -107,15 +296,32 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 rasterizationProducerStage = other.rasterizationProducerStage;
                 producerOutputComponentCount = other.producerOutputComponentCount;
                 fragmentInputComponentCount = other.fragmentInputComponentCount;
+                fragmentReplacesDepth = other.fragmentReplacesDepth;
+                readsBaseVertexBuiltin = other.readsBaseVertexBuiltin;
+                writesViewportIndexBuiltin = other.writesViewportIndexBuiltin;
+                needsPassthroughTessControl = other.needsPassthroughTessControl;
+                passthroughTessControlEmulatable = other.passthroughTessControlEmulatable;
+                passthroughPerVertexMembers = other.passthroughPerVertexMembers;
+                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;
                 other.rasterizationProducerStage = ShaderStage::Unknown;
                 other.producerOutputComponentCount = 0;
                 other.fragmentInputComponentCount = 0;
+                other.fragmentReplacesDepth = false;
+                other.readsBaseVertexBuiltin = false;
+                other.writesViewportIndexBuiltin = false;
+                other.needsPassthroughTessControl = false;
+                other.passthroughTessControlEmulatable = false;
+                other.passthroughPerVertexMembers = 0;
+                other.lastUsedFrame = 0;
             }
             VkProgramObject& operator=(VkProgramObject&& other) noexcept {
                 if (this == &other) {
@@ -125,11 +331,16 @@ 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);
+                arrayedUniformBlockIndicesByBinding = std::move(other.arrayedUniformBlockIndicesByBinding);
                 samplerNameByBinding = std::move(other.samplerNameByBinding);
                 samplerUniformLocationByBinding = std::move(other.samplerUniformLocationByBinding);
                 samplerTextureTargetByBinding = std::move(other.samplerTextureTargetByBinding);
@@ -139,6 +350,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                     std::move(other.storageImageUsesBindingFormatByBinding);
                 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;
@@ -147,15 +360,32 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 rasterizationProducerStage = other.rasterizationProducerStage;
                 producerOutputComponentCount = other.producerOutputComponentCount;
                 fragmentInputComponentCount = other.fragmentInputComponentCount;
+                fragmentReplacesDepth = other.fragmentReplacesDepth;
+                readsBaseVertexBuiltin = other.readsBaseVertexBuiltin;
+                writesViewportIndexBuiltin = other.writesViewportIndexBuiltin;
+                needsPassthroughTessControl = other.needsPassthroughTessControl;
+                passthroughTessControlEmulatable = other.passthroughTessControlEmulatable;
+                passthroughPerVertexMembers = other.passthroughPerVertexMembers;
+                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;
                 other.rasterizationProducerStage = ShaderStage::Unknown;
                 other.producerOutputComponentCount = 0;
                 other.fragmentInputComponentCount = 0;
+                other.fragmentReplacesDepth = false;
+                other.readsBaseVertexBuiltin = false;
+                other.writesViewportIndexBuiltin = false;
+                other.needsPassthroughTessControl = false;
+                other.passthroughTessControlEmulatable = false;
+                other.passthroughPerVertexMembers = 0;
+                other.lastUsedFrame = 0;
                 return *this;
             }
 
@@ -182,27 +412,171 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 }
                 modules.clear();
                 stages.clear();
+                stageSpirvDigests.clear(); // the modules they describe are gone
             }
         };
 
-        explicit ProgramFactory(VkDevice device, const VulkanRendererConfig& config, Uint32 maxBindings = 16,
-                                Bool shaderDrawParametersEnabled = false,
-                                Bool unformattedFloatStorageImagesEnabled = false)
+        // Notified when the OnFrameBoundary sweep destroys an aged-out cache entry,
+        // carrying the entry's content hash and the VkDescriptorSetLayout it owned.
+        // Dependent caches (compute pipelines, PipelineFactory entries, UniformManager's
+        // per-layout descriptor sets) must purge in the same step: after vkDestroy the
+        // layout handle value may be recycled for an unrelated layout, and the program
+        // hash may be re-inserted by a later rebuild of the same content.
+        class IEvictionObserver {
+        public:
+            virtual ~IEvictionObserver() = default;
+            virtual void OnProgramEvicted(HashType programHash, VkDescriptorSetLayout descriptorSetLayout) = 0;
+        };
+
+        // How this factory's compute modules implement GL_KHR_shader_subgroup. Computed
+        // once at renderer initialization (SubgroupSupportPolicy.h + the device's
+        // 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 tessellationAndGeometryPointSizeEnabled,
+                                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_tessellationAndGeometryPointSizeEnabled(tessellationAndGeometryPointSizeEnabled),
+              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
+        // (their command buffers retired many frames ago), mirroring
+        // VkRenderPassManager::OnPresent's sweep.
+        void OnFrameBoundary();
+
         static VkShaderStageFlagBits ToVkStage(ShaderStage stage);
         static VkFormat ConvertSpirvImageFormatToVkFormat(SpvImageFormat format);
         static SamplerNumericDomain UniformTypeToSamplerNumericDomain(GLenum glType);
+        // The same question for an IMAGE uniform (`image2D`, `uimageBuffer`, ...), which the
+        // sampler form above deliberately does not answer. Kept separate rather than folded in
+        // because the two are asked in different places for different reasons: a sampler's domain
+        // decides a sampled VIEW format, an image's decides what a placeholder descriptor for an
+        // UNBOUND image unit must be (see UniformManager::AcquireUnboundTexelBufferView and
+        // GetUnboundStorageImageTexture) - a formatless `writeonly` declaration reflects no
+        // format at all, and the numeric domain is then the only thing that constrains it.
+        static SamplerNumericDomain UniformTypeToImageNumericDomain(GLenum glType);
+        // True when any entry point declares the DepthReplacing execution mode, i.e. the
+        // shader assigns gl_FragDepth. Exposed so the blended depth-write quirk's exemption
+        // can be pinned by tests. A false negative loses the exemption, so such a shader is
+        // stripped conservatively and forfeits its depth write.
+        static Bool ReflectedFragmentReplacesDepth(const SpvReflectShaderModule& reflectModule);
+        // 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, the six default tessellation levels AND the gl_PerVertex
+        // member set, because all three decide what the generator emits. The size comes from
+        // PATCH_VERTICES and the levels from PATCH_DEFAULT_OUTER_LEVEL / PATCH_DEFAULT_INNER_LEVEL
+        // - draw state rather than link state, and the CTS case that motivated this links at the
+        // default 3 and draws at 4. The member set comes from the neighbouring evaluation stage's
+        // own SPIR-V, so two programs at different GLSL versions need different modules. The
+        // pipeline cache re-keys on the same inputs, so the module a pipeline was built with is
+        // part of that pipeline's identity. Compiling is bounded by the number of distinct
+        // (size, levels, members) combinations a program draws with - one or two in practice -
+        // and only ever happens for the rare program that has no control stage at all.
+        VkPipelineShaderStageCreateInfo GetOrCreatePassthroughTessControlStage(Uint32 patchVertices,
+                                                                              const FloatVec4& defaultOuterLevel,
+                                                                              const FloatVec2& defaultInnerLevel,
+                                                                              Uint32 perVertexMembers);
+
+        // 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, const FloatVec4& defaultOuterLevel,
+                                                        const FloatVec2& defaultInnerLevel, Uint32 perVertexMembers);
+
+        // The identity of one such module: everything the generator bakes in, folded into a
+        // 64-bit key over the raw bits (so -0.0 and +0.0 key apart, which is harmless, and NaN
+        // keys to itself, which is what matters). Shared with PipelineFactory, which mixes the
+        // same value into the pipeline hash so a pipeline can never be handed a module built for
+        // different levels or a different block shape.
+        static Uint64 ComputePassthroughTessControlKey(Uint32 patchVertices, const FloatVec4& defaultOuterLevel,
+                                                       const FloatVec2& defaultInnerLevel, Uint32 perVertexMembers);
+
+        // The PerVertexMemberBit mask of the INPUT per-vertex block a module declares, read
+        // straight out of its SPIR-V (OpMemberDecorate ... BuiltIn on the struct behind the one
+        // Input variable that is an array of a Block-decorated struct). Zero when the module has
+        // no such block. Exposed for tests, which is the only way to pin the shape agreement
+        // without a device.
+        static Uint32 ReflectPerVertexInputMembers(const Vector<Uint>& spirv);
 
     private:
         struct ProgramLookupCache {
@@ -213,14 +587,27 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         };
 
         static TextureTarget UniformTypeToTextureTarget(GLenum glType);
-        void ReflectVertexInputs(const Vector<SharedPtr<MG_State::GLState::ShaderObject>>& shaders,
+        // `stages` is ALWAYS ProgramObject::GetLinkedShaderStages() - one entry per module of
+        // `spirv`, at the same index. Taking the stages rather than the shader objects is what
+        // keeps the program's live attach list, which is a longer and differently-indexed list
+        // the moment a glAttachShader lands after the link, from being passed here by mistake.
+        void ReflectVertexInputs(const Vector<ShaderStage>& stages,
                      const Vector<Vector<Uint>>& spirv,
                      VkProgramObject& entry) const;
-        void ReflectFragmentOutputs(const Vector<SharedPtr<MG_State::GLState::ShaderObject>>& shaders,
+        void ReflectViewportIndexUsage(const Vector<ShaderStage>& stages,
+                                       const Vector<Vector<Uint>>& spirv,
+                                       VkProgramObject& entry) const;
+        void ReflectFragmentOutputs(const Vector<ShaderStage>& stages,
                         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<ShaderStage>& stages,
+                                               const Vector<Vector<Uint>>& spirv,
+                                               VkProgramObject& entry) const;
 
         VkDevice m_device = VK_NULL_HANDLE;
         Uint32 m_maxBindings = 0;
@@ -232,7 +619,42 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // True only when the logical device enabled both
         // shaderStorageImageReadWithoutFormat and shaderStorageImageWriteWithoutFormat.
         Bool m_unformattedFloatStorageImagesEnabled = false;
+        // True when the logical device enabled shaderTessellationAndGeometryPointSize. When it is
+        // FALSE a program whose tessellation or geometry module declares TessellationPointSize /
+        // GeometryPointSize is refused at build time (see VkProgramObject::
+        // pointSizeCapabilityUnsupported) instead of being handed to the driver as invalid usage.
+        Bool m_tessellationAndGeometryPointSizeEnabled = 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 the identity of what was compiled into
+        // them - the input patch size and the six default tessellation levels, folded into one
+        // 64-bit key by ComputePassthroughTessControlKey (the levels are float state, so the map
+        // cannot simply be keyed on the patch size any more). A failed build is cached as
+        // VK_NULL_HANDLE so a broken generator costs one compile, not one per draw.
+        //
+        // Hard-capped, because the key is application-controlled: glPatchParameterfv clamps
+        // nothing, so an application that recomputes a level per frame mints a new key per frame.
+        // Reaching the cap destroys every module and starts over (see the flush in
+        // GetOrCreatePassthroughTessControlStage); the cap is far above what any program that
+        // holds its levels still will ever need. The gl_PerVertex member set is in the key too
+        // and adds only a handful of values, so it does not move the cap in practice.
+        static constexpr SizeT kMaxPassthroughTessControlStages = 64;
+        UnorderedMap<Uint64, 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);
@@ -247,6 +247,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         m_surfaceFormat = {createInfo.imageFormat, createInfo.imageColorSpace};
         m_extent = createInfo.imageExtent;
+        // The surface-space extent this swapchain was built from, i.e. before the
+        // quarter-turn swap above. Out-of-date checks must compare in THIS space: comparing a
+        // freshly queried currentExtent against the swapped m_extent flips axes every rotation
+        // and makes the comparison alternate forever.
+        m_surfaceExtent = defaultFramebufferExtent;
         m_preTransform = createInfo.preTransform;
 
         VK_VERIFY(vkCreateSwapchainKHR(device, &createInfo, nullptr, &m_swapchain));
@@ -257,6 +262,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_images.resize(imageCount, VK_NULL_HANDLE);
         VK_VERIFY(vkGetSwapchainImagesKHR(device, m_swapchain, &imageCount, m_images.data()));
         m_imageLayouts.assign(imageCount, VK_IMAGE_LAYOUT_UNDEFINED);
+        // Fresh swapchain images hold garbage until a render pass stores into them.
+        m_imageContentDefined.assign(imageCount, false);
+        m_depthStencilContentDefined.assign(imageCount, false);
 
         CreateImageViews(device);
         CreateDepthStencilResources(device, physicalDevice);
@@ -428,9 +436,39 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         m_images.clear();
         m_imageLayouts.clear();
+        m_imageContentDefined.clear();
+        m_depthStencilContentDefined.clear();
         m_preTransform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
     }
 
+    Bool SwapchainObject::IsImageContentDefined(Uint32 index) const {
+        MOBILEGL_ASSERT(index < m_imageContentDefined.size(), "Swapchain image content index out of range");
+        return m_imageContentDefined[index];
+    }
+
+    void SwapchainObject::SetImageContentDefined(Uint32 index, Bool defined) {
+        MOBILEGL_ASSERT(index < m_imageContentDefined.size(), "Swapchain image content index out of range");
+        m_imageContentDefined[index] = defined;
+    }
+
+    Bool SwapchainObject::IsDepthStencilContentDefined(Uint32 index) const {
+        MOBILEGL_ASSERT(index < m_depthStencilContentDefined.size(),
+                        "Swapchain depth/stencil content index out of range");
+        return m_depthStencilContentDefined[index];
+    }
+
+    void SwapchainObject::SetDepthStencilContentDefined(Uint32 index, Bool defined) {
+        MOBILEGL_ASSERT(index < m_depthStencilContentDefined.size(),
+                        "Swapchain depth/stencil content index out of range");
+        m_depthStencilContentDefined[index] = defined;
+    }
+
+    void SwapchainObject::SetAllDepthStencilContentUndefined() {
+        for (SizeT i = 0; i < m_depthStencilContentDefined.size(); ++i) {
+            m_depthStencilContentDefined[i] = false;
+        }
+    }
+
     VkImage SwapchainObject::GetImage(Uint32 index) const {
         MOBILEGL_ASSERT(index < m_images.size(), "Swapchain image index out of range");
         return m_images[index];

MobileGL/MG_Backend/DirectVulkan/Renderer/SwapchainObject.h

@@ -35,6 +35,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkSwapchainKHR GetHandle() const { return m_swapchain; }
         const VkSurfaceFormatKHR& GetSurfaceFormat() const { return m_surfaceFormat; }
         VkExtent2D GetExtent() const { return m_extent; }
+        // Surface-space extent (before the pre-rotation quarter-turn swap) this swapchain was
+        // created from - the value to compare a freshly queried currentExtent against.
+        VkExtent2D GetSurfaceExtent() const { return m_surfaceExtent; }
         VkSurfaceTransformFlagBitsKHR GetPreTransform() const { return m_preTransform; }
         const Vector<VkImage>& GetImages() const { return m_images; }
         const Vector<VkImageView>& GetImageViews() const { return m_imageViews; }
@@ -49,6 +52,21 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         void SetImageLayout(Uint32 index, VkImageLayout layout);
         SizeT GetImageCount() const { return m_images.size(); }
 
+        // EGL content-validity tracking for the default framebuffer. A color
+        // buffer's content is undefined once its image has been presented
+        // (EGL_BUFFER_DESTROYED swap behaviour, the implementation default),
+        // and every ancillary (depth/stencil) buffer's content is undefined
+        // after ANY swap regardless of swap behaviour (EGL 1.5 §3.10.1). The
+        // render-pass manager turns an undefined attachment's tile load into
+        // LOAD_OP_DONT_CARE. Flags start false (a fresh swapchain image holds
+        // garbage) and a render pass storing into an attachment sets it back
+        // to defined.
+        Bool IsImageContentDefined(Uint32 index) const;
+        void SetImageContentDefined(Uint32 index, Bool defined);
+        Bool IsDepthStencilContentDefined(Uint32 index) const;
+        void SetDepthStencilContentDefined(Uint32 index, Bool defined);
+        void SetAllDepthStencilContentUndefined();
+
     private:
         void CreateImageViews(VkDevice device);
         void CreateDepthStencilResources(VkDevice device, VkPhysicalDevice physicalDevice);
@@ -63,6 +81,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkSwapchainKHR m_swapchain = VK_NULL_HANDLE;
         VkSurfaceFormatKHR m_surfaceFormat{};
         VkExtent2D m_extent{};
+        VkExtent2D m_surfaceExtent{};
         VkSurfaceTransformFlagBitsKHR m_preTransform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
         Vector<VkImage> m_images;
         Vector<VkImageView> m_imageViews;
@@ -73,5 +92,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Vector<VkDeviceMemory> m_depthStencilImageMemories;
         Vector<VkImageView> m_depthStencilImageViews;
         Vector<VkImageLayout> m_depthStencilImageLayouts;
+        Vector<Bool> m_imageContentDefined;
+        Vector<Bool> m_depthStencilContentDefined;
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/UniformManager.h

@@ -26,12 +26,26 @@ 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;
         };
 
-        Bool Initialize(VkDevice device, VkBufferManager* bufferManager,
+        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;
+        };
+
+        // `physicalDevice` is only ever asked for format properties: a placeholder descriptor for
+        // an unbound texel-buffer binding has to be built from a format the DEVICE accepts as a
+        // texel buffer, and there is no other route to that answer from here.
+        Bool Initialize(VkDevice device, VkPhysicalDevice physicalDevice, VkBufferManager* bufferManager,
                         ProgramFactory* programFactory,
                         VkDeviceSize minUniformBufferOffsetAlignment, Uint32 frameCount,
                         Uint32 maxBindings = 16, Uint32 setsPerFrame = 64,
@@ -39,18 +53,66 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         void Shutdown();
 
         void BeginFrame(Uint32 frameIndex);
+        // A command buffer (re)began recording: descriptor bindings recorded into
+        // the previous buffer do not carry over, so drop the bind-dedup shadow.
+        void OnCommandBufferBoundary() { m_lastBindValid = false; }
+        // A ProgramFactory eviction just destroyed this layout: purge every frame
+        // slot's cached descriptor sets for it, so a recycled handle value can never
+        // stale-hit sets written for the dead layout's bindings. The sets are
+        // vkFreeDescriptorSets'd back to their pools (created with
+        // FREE_DESCRIPTOR_SET_BIT) and the pool accounting is credited, so program
+        // churn recycles pool capacity instead of abandoning it. GPU-safe: the layout
+        // only dies after >1024 idle frame boundaries, so no in-flight command buffer
+        // references its sets. This is the only eviction path for the per-layout
+        // 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
@@ -58,15 +120,34 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         static VkFormat ResolveStorageImageViewFormat(VkFormat reflectedFormat, GLenum bindingFormat,
                                                       VkFormat resourceFormat, Bool useBindingFormat);
 
+        // True when the program reads at least one sampler and every one of them is bound to a
+        // texture whose GL level range is a single level. Such a sampler resolves to
+        // minLod = maxLod = 0 (see VkSamplerManager::GetOrCreateSampler), so an implicit-LOD sample
+        // and an explicit LOD 0 sample must read the same texel - which is what makes the
+        // ExplicitLod0Sampling SPIR-V rewrite safe to request. Deliberately conservative: it reads
+        // only GL state, so a texture that ends up single-level for another reason (one uploaded
+        // level under a wide level range) merely misses the rewrite.
+        static Bool ProgramSamplesOnlySingleLevelTextures(const MG_State::GLState::ProgramObject& program,
+                                                          const ProgramFactory::VkProgramObject& programObj);
+
     private:
         struct DescriptorPoolBucket {
             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
+        // layout-destroyed purge can vkFreeDescriptorSets it back and credit the
+        // owning bucket's accounting.
+        struct CachedDescriptorSet {
+            VkDescriptorSet set = VK_NULL_HANDLE;
+            VkDescriptorPool pool = VK_NULL_HANDLE;
         };
 
         struct DescriptorSetCacheEntry {
-            Vector<VkDescriptorSet> sets;
+            Vector<CachedDescriptorSet> sets;
             Uint32 cursor = 0;
         };
 
@@ -82,28 +163,96 @@ 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);
-        SharedPtr<MG_State::GLState::ITextureObject> GetFallbackTexture(TextureTarget target) const;
+            const ProgramFactory::VkProgramObject& programObj, Uint32 binding, Uint32 element);
+        // `numericDomain` is the sampler's class, and it matters only for the multisample arm -
+        // see GetFallbackMultisampleTexture for why the single-sampled fallback can ignore it.
+        SharedPtr<MG_State::GLState::ITextureObject> GetFallbackTexture(
+            TextureTarget target, SamplerNumericDomain numericDomain) const;
+        // The multisample arm of GetFallbackTexture. One object per (target, numeric domain) and
+        // no upload path: a multisample image cannot be written by a transfer, so its texels stay
+        // undefined - which is what GL promises for a texelFetch on an incomplete multisample
+        // texture - and it cannot carry MUTABLE_FORMAT, so its format has to match the sampler's
+        // class outright rather than being reinterpreted at view time.
+        SharedPtr<MG_State::GLState::ITextureObject> GetFallbackMultisampleTexture(
+            TextureTarget target, SamplerNumericDomain numericDomain) const;
+        // ---- placeholders for UNBOUND image-backed descriptors -------------------------
+        // GL lets a program declare `samplerBuffer`, `imageBuffer` or `image2D` and bind nothing
+        // to the unit it names: the fetch is then undefined (GL 4.6 core 8.9 for an incomplete
+        // buffer texture, 8.26 for an image unit with no texture) - undefined VALUES, not a
+        // dropped draw. Vulkan has no unwritten descriptor, so something valid has to sit in the
+        // set or the whole draw or dispatch is lost, which is what these two build. Same shape as
+        // VkBufferManager::AcquireUnboundStorageDescriptor, one level up: per FORMAT rather than
+        // one shared object, because a descriptor whose format disagrees with the shader's
+        // declaration is invalid Vulkan even when nothing ever reads it.
+        //
+        // `declaredFormat` is the format the SHADER declared (VK_FORMAT_UNDEFINED for a sampled
+        // texel buffer, which never carries one, or for a formatless `writeonly` image);
+        // `numericDomain` decides the format when there is no declaration and is the fallback
+        // class when the device cannot use the declared one as a texel buffer.
+        VkBufferView AcquireUnboundTexelBufferView(VkFormat declaredFormat, SamplerNumericDomain numericDomain,
+                                                   Bool storage);
+        // A 1x1 (x1 layer, or 6 faces for a cube) texture of `format`, shaped for `target` so the
+        // view the descriptor gets has the view type the shader's image declaration demands.
+        // Null for a target with no single-sampled placeholder shape - multisample images, whose
+        // descriptor needs a multisample view that this cannot stand in for.
+        SharedPtr<MG_State::GLState::ITextureObject> GetUnboundStorageImageTexture(TextureTarget target,
+                                                                                    VkFormat format) const;
+        // The (target, format) pair a storage-image binding's placeholder is keyed by, resolved
+        // from reflection alone. False when the binding has no placeholder shape.
+        Bool ResolveUnboundStorageImagePlaceholder(const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
+                                                   TextureTarget& outTarget, VkFormat& outFormat) 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 {
@@ -116,9 +265,24 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         };
         Bool ResolveUniformBufferPayload(const MG_State::GLState::ProgramObject& program,
                                          const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
-                                         UboBindResult& out) const;
-        Bool CreateDescriptorPool(Uint32 maxSets, VkDescriptorPool& outPool) const;
-        Bool GrowFrameDescriptorPool(FrameResources& frame, Uint32 frameIndex);
+                                         Uint32 arrayElement, UboBindResult& out) const;
+        // Shared resolution of one dynamic-UBO binding element into the
+        // (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,
@@ -126,6 +290,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                       VkDescriptorSet& outDescriptorSet);
 
         VkDevice m_device = VK_NULL_HANDLE;
+        VkPhysicalDevice m_physicalDevice = VK_NULL_HANDLE;
         VkBufferManager* m_bufferManager = nullptr;
         ProgramFactory* m_programFactory = nullptr;
         Vector<FrameResources> m_frames;
@@ -138,6 +303,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkTextureManager* m_textureManager = nullptr;
         VkSamplerManager* m_samplerManager = nullptr;
         mutable SharedPtr<MG_State::GLState::ITextureObject> m_fallbackTexture2D;
+        // Keyed by (arrayed, numeric domain); see GetFallbackMultisampleTexture. Lazily populated,
+        // never evicted - at most six tiny 1x1 images - and torn down with the manager.
+        mutable UnorderedMap<Uint32, SharedPtr<MG_State::GLState::ITextureObject>> m_fallbackMultisampleTextures;
+        // See AcquireUnboundTexelBufferView / GetUnboundStorageImageTexture. Both are lazily
+        // populated, never evicted (a program's declared formats are a fixed, tiny set) and torn
+        // down with the manager. The texel views are keyed by format AND by storage-vs-sampled
+        // because the two descriptor kinds demand different format FEATURES of the device, so one
+        // format can be usable for one and not the other. Deliberately NOT the per-frame
+        // texelBufferViews list: those are destroyed at every frame boundary, and these must
+        // outlive it or the placeholder would be rebuilt for every unbound binding every frame.
+        UnorderedMap<Uint64, VkBufferView> m_unboundTexelBufferViews;
+        mutable UnorderedMap<Uint64, SharedPtr<MG_State::GLState::ITextureObject>> m_unboundStorageImageTextures;
 
         // Per-draw scratch buffers for BindProgramUniformBuffers: reused (clear keeps
         // capacity) so the descriptor-write path stops allocating on every draw.
@@ -147,14 +324,82 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Vector<VkBufferView> m_texelBufferViewsScratch;
         Vector<Uint32> m_dynamicOffsetsScratch;
 
-        // Descriptor-set reuse across consecutive draws (see BindProgramUniformBuffers).
-        // When a draw's resolved descriptor content is byte-identical to the previous
-        // draw's, reuse the same VkDescriptorSet and skip AcquireDescriptorSet +
-        // vkUpdateDescriptorSets - only the bind-time dynamic offsets differ. Reset each
-        // frame in BeginFrame because the frame's descriptor sets are recycled there.
-        VkDescriptorSet m_lastBoundDescriptorSet = VK_NULL_HANDLE;
-        Uint64 m_lastDescriptorSignature = 0;
-        Bool m_hasLastDescriptor = false;
+        // Descriptor-set reuse across recent draws (see BindProgramUniformBuffers).
+        // When a draw's resolved descriptor content is byte-identical to one memoized
+        // earlier, reuse that VkDescriptorSet and skip AcquireDescriptorSet +
+        // vkUpdateDescriptorSets - only the bind-time dynamic offsets differ. Four
+        // entries with round-robin replacement rather than one: draws alternating
+        // between two programs (MC's chunk<->entity ping-pong) would thrash a single
+        // slot into a full re-allocate+write every draw. Reset each frame in BeginFrame
+        // 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
+        // driver call can be skipped outright. Command-buffer-scope state; reset
+        // via OnCommandBufferBoundary whenever a recording (re)begins. Keyed on
+        // layout+bind point, so a pipeline-layout switch always rebinds.
+        static constexpr Uint32 kMaxShadowedDynamicOffsets = 8;
+        Bool m_lastBindValid = false;
+        VkDescriptorSet m_lastBindSet = VK_NULL_HANDLE;
+        VkPipelineLayout m_lastBindLayout = VK_NULL_HANDLE;
+        VkPipelineBindPoint m_lastBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
+        Uint32 m_lastBindOffsetCount = 0;
+        Uint32 m_lastBindOffsets[kMaxShadowedDynamicOffsets] = {};
+
+        // Global-UBO transient-slice reuse: MC leaves the default uniform block
+        // untouched across long GUI/terrain runs, so the per-draw re-upload of
+        // the same bytes can reuse the slice uploaded earlier THIS frame (frame
+        // serial guards arena recycling; the content version guards writes).
+        struct GlobalUboSliceMemo {
+            Uint64 programLifetimeId = 0;
+            Uint64 frameSerial = 0;
+            Uint32 uboContentVersion = 0;
+            VkBuffer buffer = VK_NULL_HANDLE;
+            VkDeviceSize offset = 0;
+            VkDeviceSize range = 0;
+        };
+        static constexpr Uint32 kGlobalUboMemoSize = 4;
+        GlobalUboSliceMemo m_globalUboMemo[kGlobalUboMemoSize];
+        Uint32 m_globalUboMemoNext = 0;
 
         // Per-binding fast path over VkSamplerManager's content-hashed sampler cache, which
         // stays the source of truth: its key hashes all sampler+texture state, so two distinct
@@ -167,16 +412,56 @@ 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;
+            Uint32 viewLevelCount = 0;
             Uint16 samplerVersion = 0;
             Uint16 textureParamsVersion = 0;
             Bool forceNearestFiltering = false;
             Bool valid = false;
+            // ResolveSampledImageViewFormat is pure in (image format, numeric domain), but a
+            // domain mismatch walks a ~184-entry format table. Memo the resolution per binding
+            // so a reinterpreted sampler pays that scan once, not once per draw.
+            VkFormat viewFormatSource = VK_FORMAT_UNDEFINED;
+            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

@@ -8,6 +8,7 @@
 
 #include "VertexInputStateFactory.h"
 #include "MG_Util/Converters/MGToStr/DataTypeConverter.h"
+#include <MG_Backend/BackendObjects.h>
 #include <utility>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
@@ -29,10 +30,22 @@ 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)));
 
-            const SizeT bufferKey = reinterpret_cast<SizeT>(attr.Buffer.get());
+            // The bound buffer's IDENTITY is a component of the key, and it has to be the
+            // buffer's never-reused lifetime id - NOT its heap address, which this used to
+            // hash. An address is recycled by the allocator, so a deleted-and-recreated
+            // buffer reproduces it; combined with a byte-identical attribute layout that
+            // reproduces the WHOLE content hash, and the hash is what
+            // TryBindResolvedVertexBindings accepts as proof that a memoised binding still
+            // reads the buffer it was resolved from. It did not: a destroyed buffer's GPU
+            // 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)));
         }
 
@@ -51,14 +64,33 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
     const VertexInputStateFactory::BackendVertexInputState& VertexInputStateFactory::GetOrCreateVertexInputState(
         const MG_State::GLState::VertexArrayObject& vao) {
-        return GetOrCreateVertexInputState(vao, GetOrComputeHash(vao));
+        // Per-draw fast path: the VAO carries a pointer to its resolved entry,
+        // valid while its config version and the cache's eviction epoch both
+        // match - no re-hash, no map lookup.
+        const void* memoState = nullptr;
+        Uint64 memoEpoch = 0;
+        if (vao.GetBackendStateMemo(memoState, memoEpoch) && memoEpoch == m_evictionEpoch) {
+            const auto* entry = static_cast<const BackendVertexInputState*>(memoState);
+            entry->lastUsedFrameBoundary = m_frameBoundaryCounter;
+            return *entry;
+        }
+        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;
     }
 
     const VertexInputStateFactory::BackendVertexInputState& VertexInputStateFactory::GetOrCreateVertexInputState(
         const MG_State::GLState::VertexArrayObject& vao, HashType hash) {
         auto it = m_cache.find(hash);
         if (it != m_cache.end()) {
-            return it->second;
+            it->second->lastUsedFrameBoundary = m_frameBoundaryCounter;
+            return *it->second;
         }
 
         VertexInputStateBuilder builder;
@@ -67,6 +99,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Vector<Uint32> bindingAttributeLocations;
         Vector<Bool> bindingUsesClientMemory;
         Vector<VertexStreamConversion> bindingConversions;
+        Vector<VkVertexInputBindingDivisorDescriptionEXT> bindingDivisors;
         Uint32 unsupportedAttribMask = 0;
 
         for (Uint32 location = 0; location < MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS; ++location) {
@@ -75,10 +108,36 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 continue;
             }
 
-            const VkFormat sourceVkFormat =
-                ToVkVertexFormat(attr.Type, attr.Size, attr.Normalized, attr.IsInteger, attr.IsBgra);
+            VkFormat sourceVkFormat =
+                ToVkVertexFormat(attr.Type, attr.Size, attr.Normalized, attr.IsInteger, attr.IsBgra, attr.IsLong);
+            VertexStreamConversion conversion = VertexStreamConversion::None;
+            // Gated on the SAME flag ToVkVertexFormat gates its 64-bit path on, and that is
+            // load-bearing rather than belt-and-braces: the narrowing is only correct because the
+            // shader's `dvec` input is a `vec` by the time the pipeline is built, and what
+            // guarantees that is the flag being clear. It is clear on every backend today, and a
+            // program with a 64-bit float vertex input is demoted WHOLE for the same reason even
+            // where the device has native fp64 (ProgramSpirvTask::GenerateSpirv). With the flag
+            // set, a dvec3/dvec4 would be declined by ToVkVertexFormat AND left 64-bit in the
+            // module, so a float32 stream would be fed to a Float64 input.
+            const Bool narrowFloat64Arrays =
+                MG_Backend::pActiveBackendObject == nullptr ||
+                !MG_Backend::pActiveBackendObject->GetDynamicParameters().SupportsFloat64VertexAttributes;
+            if (sourceVkFormat == VK_FORMAT_UNDEFINED && attr.Type == DataType::Float64 && narrowFloat64Arrays) {
+                // No native 64-bit fetch here (see ToVkVertexFormat's Float64 case), but the
+                // source bytes are ordinary IEEE-754 doubles and DemoteFloat64Pass has already
+                // narrowed every dvec input to a vec, so the array is narrowed to match rather
+                // than dropped. Mirrors what DirectGLES does for the same state.
+                const VkFormat narrowedFormat = ToFloat32VertexFormat(attr.Size);
+                if (narrowedFormat != VK_FORMAT_UNDEFINED && SupportsVertexBufferFormat(narrowedFormat)) {
+                    sourceVkFormat = narrowedFormat;
+                    conversion = VertexStreamConversion::Float64ToFloat32;
+                    MGLOG_W_ONCE("Vertex attribute location=%u is a 64-bit (GL_DOUBLE) array; fetching it at "
+                            "float32 precision through format=%d (size=%d long=%s)",
+                            location, static_cast<Int>(narrowedFormat), attr.Size, attr.IsLong ? "true" : "false");
+                }
+            }
             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);
@@ -86,14 +145,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             }
 
             VkFormat vkFormat = sourceVkFormat;
-            VertexStreamConversion conversion = VertexStreamConversion::None;
-            if (!SupportsVertexBufferFormat(vkFormat)) {
+            if (conversion == VertexStreamConversion::None && !SupportsVertexBufferFormat(vkFormat)) {
                 if (IsScaledIntegerVertexFormat(vkFormat)) {
                     const VkFormat fallbackFormat = ToFloat32VertexFormat(attr.Size);
                     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),
@@ -103,7 +161,7 @@ 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);
@@ -114,34 +172,52 @@ 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 =
-                attr.Stride > 0 ? static_cast<Uint32>(attr.Stride) : static_cast<Uint32>(attribByteSize);
+            // Verbatim, zero included. The frontend already resolved a pointer call's
+            // "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);
+            // For a client-memory array attr.Offset holds the raw client pointer, and the
+            // draw path re-uploads the data to a 16-aligned transient slice with attribute
+            // offset 0, so only the stride can violate Vulkan's fetch alignment there.
+            const Bool clientMemoryAttribute = attr.Buffer == nullptr;
             if (conversion == VertexStreamConversion::None && requiredAlignment > 1 &&
-                ((sourceStride % requiredAlignment) != 0 || (attr.Offset % requiredAlignment) != 0)) {
+                ((sourceStride % requiredAlignment) != 0 ||
+                 (!clientMemoryAttribute && (attr.Offset % requiredAlignment) != 0))) {
                 // GL accepts arbitrary byte strides and offsets. Core Vulkan vertex fetches do not
                 // 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;
-            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)));
+            // 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 ||
+                           conversion == VertexStreamConversion::Float64ToFloat32) {
+                    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;
@@ -155,14 +231,51 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             bindingConversions.push_back(conversion);
             builder.AddBinding(binding, stride, inputRate);
             builder.AddAttribute(location, binding, vkFormat, 0);
+            // Divisor 1 is what VK_VERTEX_INPUT_RATE_INSTANCE already means; only anything
+            // else needs the extension to say it.
+            if (inputRate == VK_VERTEX_INPUT_RATE_INSTANCE && attr.Divisor != 1) {
+                bindingDivisors.push_back({binding, static_cast<Uint32>(attr.Divisor)});
+            }
         }
 
         const auto& state = builder.Build();
 
-        auto& entry = m_cache[hash];
+        auto& slot = m_cache[hash];
+        if (!slot) {
+            slot = MakeUnique<BackendVertexInputState>();
+        }
+        BackendVertexInputState& entry = *slot;
         entry.hash = hash;
+        entry.lastUsedFrameBoundary = m_frameBoundaryCounter;
+        entry.bindingDivisors = Move(bindingDivisors);
         entry.bindings = builder.GetBindings();
         entry.attributes = builder.GetAttributes();
+        // See the layoutHash declaration: hash only the resolved layout, never
+        // buffer identities, so identical layouts across VAOs/buffers agree.
+        XXHASH_VERIFY(XXH64_reset(m_hashState, 0));
+        for (const auto& binding : entry.bindings) {
+            XXHASH_VERIFY(XXH64_update(m_hashState, &binding.binding, sizeof(binding.binding)));
+            XXHASH_VERIFY(XXH64_update(m_hashState, &binding.stride, sizeof(binding.stride)));
+            XXHASH_VERIFY(XXH64_update(m_hashState, &binding.inputRate, sizeof(binding.inputRate)));
+        }
+        for (const auto& attribute : entry.attributes) {
+            XXHASH_VERIFY(XXH64_update(m_hashState, &attribute.location, sizeof(attribute.location)));
+            XXHASH_VERIFY(XXH64_update(m_hashState, &attribute.binding, sizeof(attribute.binding)));
+            XXHASH_VERIFY(XXH64_update(m_hashState, &attribute.format, sizeof(attribute.format)));
+            XXHASH_VERIFY(XXH64_update(m_hashState, &attribute.offset, sizeof(attribute.offset)));
+        }
+        for (const auto& divisor : entry.bindingDivisors) {
+            XXHASH_VERIFY(XXH64_update(m_hashState, &divisor.binding, sizeof(divisor.binding)));
+            XXHASH_VERIFY(XXH64_update(m_hashState, &divisor.divisor, sizeof(divisor.divisor)));
+        }
+        XXHASH_VERIFY(XXH64_update(m_hashState, &unsupportedAttribMask, sizeof(unsupportedAttribMask)));
+        entry.layoutHash = XXH64_digest(m_hashState);
+        entry.attributeLocationMask = 0;
+        for (const auto& attribute : entry.attributes) {
+            if (attribute.location < 32u) {
+                entry.attributeLocationMask |= (1u << attribute.location);
+            }
+        }
         entry.bindingBufferKeys = std::move(bindingBufferKeys);
         entry.bindingBaseOffsets = std::move(bindingBaseOffsets);
         entry.bindingAttributeLocations = std::move(bindingAttributeLocations);
@@ -172,11 +285,47 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         entry.state = state;
         entry.state.pVertexBindingDescriptions = entry.bindings.empty() ? nullptr : entry.bindings.data();
         entry.state.pVertexAttributeDescriptions = entry.attributes.empty() ? nullptr : entry.attributes.data();
+        if (!entry.bindingDivisors.empty()) {
+            entry.divisorState.vertexBindingDivisorCount = static_cast<Uint32>(entry.bindingDivisors.size());
+            entry.divisorState.pVertexBindingDivisors = entry.bindingDivisors.data();
+            entry.state.pNext = &entry.divisorState;
+        } else {
+            entry.state.pNext = nullptr;
+        }
         return entry;
     }
 
+    void VertexInputStateFactory::OnFrameBoundary() {
+        ++m_frameBoundaryCounter;
+
+        // Sweep occasionally; evict entries whose last hit is far in the past.
+        // Erasure happens only here, never mid-frame: the draw path holds a
+        // reference into the current entry across its setup, and unordered_map
+        // erase would invalidate it. Entries are CPU-side only, so no GPU-idle
+        // proof is needed; an evicted entry that is used again is simply rebuilt
+        // from the VAO state (same hash, same content).
+        constexpr Uint64 kSweepInterval = 256;
+        constexpr Uint64 kRetireAgeBoundaries = 1024;
+        if ((m_frameBoundaryCounter % kSweepInterval) != 0) {
+            return;
+        }
+
+        for (auto it = m_cache.begin(); it != m_cache.end();) {
+            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. Advance through the
+                // process-wide source so the value stays unique across factory
+                // instances (see the member comment).
+                m_evictionEpoch = ++s_evictionEpochSource;
+            } else {
+                ++it;
+            }
+        }
+    }
+
     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
@@ -201,6 +350,36 @@ 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.
+            //
+            // ... as long as the shader half still runs. It does not when the backend has declared
+            // no 64-bit vertex attribute support: DemoteFloat64Pass has already narrowed every
+            // `dvec` input to a `vec` by then, so PackDoubleVertexInputsPass finds nothing to pack
+            // and a UINT-formatted attribute would be fed to a float input - garbage with no
+            // diagnostic anywhere. Declining here hands the attribute to the caller's
+            // Float64ToFloat32 fallback instead, which narrows the source doubles to match the
+            // demoted `vec` input - the same thing DirectGLES does for the same state. The
+            // frontend RECORDS the format either way, so this gate is the only thing standing
+            // between a legal glVertexAttribLFormat and a mismatched pipeline.
+            if (MG_Backend::pActiveBackendObject == nullptr ||
+                !MG_Backend::pActiveBackendObject->GetDynamicParameters().SupportsFloat64VertexAttributes) {
+                return VK_FORMAT_UNDEFINED;
+            }
+            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

@@ -23,10 +23,26 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             None = 0,
             Repack,
             ScaledIntegerToFloat32,
+            // GL_DOUBLE source data narrowed to a tightly packed float32 stream: the fetch half
+            // of the fp64 demotion the shader side already does unconditionally.
+            Float64ToFloat32,
         };
 
         struct BackendVertexInputState {
             HashType hash = 0;
+            // Hash of the resolved Vulkan vertex layout only (bindings, attributes,
+            // unsupported mask) - NO buffer identities. `hash` mixes each bound
+            // buffer's never-reused LIFETIME ID, so per-chunk VBOs mint a fresh
+            // identity per buffer; keying pipelines on that minted one VkPipeline per
+            // chunk section for an identical layout, defeating pipeline reuse and the
+            // 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).
+            // Mutable: the VAO's state-pointer memo fast path stamps it through
+            // a const entry reference.
+            mutable Uint64 lastUsedFrameBoundary = 0;
             Vector<VkVertexInputBindingDescription> bindings;
             Vector<VkVertexInputAttributeDescription> attributes;
             Vector<SizeT> bindingBufferKeys;
@@ -38,6 +54,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             // absent from `attributes`, so without this mask the draw path cannot tell them apart from
             // a genuinely disabled array and would silently feed the shader the current attribute value.
             Uint32 unsupportedAttribMask = 0;
+            // Bitmask of `attributes[i].location` - the draw path needs it up to
+            // three times per draw, so it is baked once at build time.
+            Uint32 attributeLocationMask = 0;
+            // Per-binding glVertexAttribDivisor values other than 1. Vulkan's instance input
+            // rate advances once per instance and nothing else, so anything else has to be
+            // stated through VK_EXT_vertex_attribute_divisor. Empty when every instanced
+            // binding uses divisor 1, which is what the plain input rate already means.
+            Vector<VkVertexInputBindingDivisorDescriptionEXT> bindingDivisors;
+            VkPipelineVertexInputDivisorStateCreateInfoEXT divisorState{
+                VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_DIVISOR_STATE_CREATE_INFO_EXT
+            };
             VkPipelineVertexInputStateCreateInfo state{
                 VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO
             };
@@ -48,6 +75,13 @@ 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.
@@ -55,6 +89,16 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         const BackendVertexInputState& GetOrCreateVertexInputState(
             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 each bound buffer's never-reused lifetime id, so
+        // 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
+        // compare except on sweep boundaries.
+        void OnFrameBoundary();
         static SizeT GetComponentSize(DataType type);
         // Tightly-packed byte size of one vertex element for this attribute: componentSize * size for
         // normal types, and 4 (one packed word) for the 2_10_10_10 types and GL_BGRA. Returns 0 for
@@ -62,14 +106,39 @@ 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;
-        UnorderedMap<HashType, BackendVertexInputState> m_cache;
+        // Values are heap-allocated: UnorderedMap is open-addressing, so INSERT
+        // invalidates references to stored values - and so does ERASE, which shifts
+        // the rest of the probe cluster into the hole and therefore moves entries
+        // 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;
+        // Bumped whenever any cache entry is erased. VAOs memo a raw pointer to
+        // their heap-allocated entry (stable across map insert/rehash by
+        // construction); a memo is honored only while its recorded epoch
+        // matches, so an evicted entry can never be dereferenced through a
+        // stale memo.
+        //
+        // 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

@@ -7,6 +7,8 @@
 // End of Source File Header
 
 #include "VkBufferManager.h"
+#include "../DirectVulkan.h"
+#include "VulkanRenderer.h"
 
 namespace MobileGL::MG_Backend::DirectVulkan {
     namespace {
@@ -14,6 +16,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
         constexpr SizeT kLiveResourcePruneThreshold = 256;
 
+        // See VkBufferManager::AcquireUnboundStorageDescriptor. 256 bytes: comfortably past
+        // every minStorageBufferOffsetAlignment in the wild, and free.
+        constexpr VkDeviceSize kUnboundStorageDescriptorBytes = 256;
+        // See VkBufferManager::AcquireUnboundTexelBufferDescriptor. The same 256 bytes, for the
+        // same reason plus one: a texel buffer view's range must be a whole number of texels of
+        // whatever format the placeholder is asked for, and 256 divides by every texel size in
+        // the GL image-format table (1, 2, 4, 8 and 16 bytes).
+        constexpr VkDeviceSize kUnboundTexelBufferDescriptorBytes = 256;
+
         // A zero-copy persistent buffer is created once and never recreated (the app holds
         // its mapped pointer), and may be bound to any role, so it carries every usage.
         // TRANSFER_DST is added by CreateResidentStorage.
@@ -21,7 +32,15 @@ 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 =
+            VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT;
         // The app writes into the persistent map with no explicit flush, so its memory must
         // be host-coherent (Adreno host-visible memory is; requiring it keeps us portable).
         constexpr VkMemoryPropertyFlags kPersistentBackedRequiredFlags =
@@ -53,6 +72,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             }
         }
 
+        // The CPU is about to read a buffer a shader wrote. Its bytes live in coherent
+        // host-visible GPU storage (EnsureGpuResidentStorage adopts it when the buffer is
+        // bound as a shader storage buffer), so nothing needs copying - but coherence only
+        // says the writes are visible once they have happened, so the work has to retire
+        // first.
+        void Ops_ReadbackFromGpu(BufferObject& bufferObject) {
+            (void)bufferObject;
+            if (pVulkanRenderer) {
+                pVulkanRenderer->FinishPendingGpuWork();
+            }
+        }
+
         void* Ops_AcquirePersistentMap(BufferObject& bufferObject) {
             if (g_activeBufferManager) {
                 return g_activeBufferManager->AcquirePersistentMap(bufferObject);
@@ -76,6 +107,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             .FlushMappedRange = Ops_FlushMappedRange,
             .OnDestroy = Ops_OnDestroy,
             .AcquirePersistentMap = Ops_AcquirePersistentMap,
+            .ReadbackFromGpu = Ops_ReadbackFromGpu,
         };
     } // namespace
 
@@ -107,6 +139,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             }
         }
         m_transientUploadArena.Shutdown();
+        m_unboundStorageBuffer.Destroy();
+        m_unboundTexelBuffer.Destroy();
         DestroyAllDeferredReleases();
         ReleaseAllLiveResources();
         m_copyProvider = nullptr;
@@ -141,6 +175,26 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_transientUploadArena.BeginFrame(frameIndex);
     }
 
+    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);
+        }
+    }
+
     void VkBufferManager::NotifyDeviceIdle() {
         // Everything submitted so far has completed. Work recorded for the
         // current frame has not been submitted yet, so the current serial
@@ -214,8 +268,15 @@ 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);
     }
@@ -223,6 +284,7 @@ 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;
@@ -237,6 +299,9 @@ 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({
@@ -248,7 +313,7 @@ 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;
@@ -270,7 +335,7 @@ 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;
         }
@@ -323,8 +388,18 @@ 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
         }
@@ -345,7 +420,7 @@ 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;
         }
     }
@@ -355,6 +430,9 @@ 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;
@@ -367,7 +445,7 @@ 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;
@@ -387,6 +465,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (!resource) {
             return;
         }
+        BumpSliceEpoch(*resource);
         resource->transientFrameSerial = 0;
         if (!resource->buffer.IsValid() || resource->pendingFullUpload) {
             return;
@@ -403,7 +482,7 @@ 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;
@@ -446,6 +525,13 @@ 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();
@@ -456,7 +542,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // it from the current shadow - MappedData() is still the shadow here because the
         // frontend adopts (and drops) the shadow only after this returns.
         DeferRelease(std::move(resource->buffer));
-        if (!CreateResidentStorage(*resource, size, kPersistentBackedUsage, kPersistentBackedRequiredFlags)) {
+        const VkBufferUsageFlags persistentUsage =
+            kPersistentBackedUsage |
+            (m_initInfo.transformFeedbackUsageEnabled ? kTransformFeedbackUsage : 0);
+        if (!CreateResidentStorage(*resource, size, persistentUsage, kPersistentBackedRequiredFlags)) {
             resource->persistentMapped = false;
             resource->storageSize = 0;
             resource->usageFlags = 0;
@@ -485,7 +574,7 @@ 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;
         }
 
@@ -507,7 +596,7 @@ 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;
@@ -530,9 +619,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         auto resource = GetOrCreateResource(bufferObject);
         bufferObject->SyncPersistentMappedRange();
 
+        // A persistently mapped resource's storage IS the application's copy of the bytes -
+        // the frontend adopted it in place of the shadow and hands out pointers into it, and
+        // a shader can have written bytes the shadow never saw (a transform feedback
+        // capture). Streaming a second copy would feed this draw the stale shadow, and the
+        // downgrade below would release the storage the application still points at,
+        // breaking the "never recreated" promise AcquirePersistentMap makes.
+        if (resource->persistentMapped) {
+            return AcquireResidentSlice(kind, bufferObject, outSlice);
+        }
+
         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;
         }
 
@@ -543,6 +642,41 @@ 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;
@@ -583,6 +717,70 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_deferredResourceReleases[frameIndex].clear();
     }
 
+    BufferSlice VkBufferManager::AcquireUnboundStorageDescriptor() {
+        if (!m_unboundStorageBuffer.IsValid()) {
+            if (m_initInfo.allocator == nullptr) {
+                return {};
+            }
+            // Host-visible so the zero fill needs no command buffer: this can be reached from
+            // descriptor resolution, which runs inside an already-open recording and must not
+            // start a copy of its own. The size is a whole minStorageBufferOffsetAlignment-safe
+            // block rather than 4 bytes so that a shader which does read the block gets a
+            // plausible unsized-array length instead of one that rounds to zero.
+            const Bool created = m_unboundStorageBuffer.Create({
+                .allocator = m_initInfo.allocator,
+                .size = kUnboundStorageDescriptorBytes,
+                .usage = VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
+                .memoryUsage = VMA_MEMORY_USAGE_AUTO,
+                .allocationFlags = VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT |
+                                   VMA_ALLOCATION_CREATE_MAPPED_BIT,
+                .requiredFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
+            });
+            if (!created) {
+                MGLOG_E_ONCE("VkBufferManager::AcquireUnboundStorageDescriptor: placeholder creation failed");
+                m_unboundStorageBuffer.Destroy();
+                return {};
+            }
+            if (void* mapped = m_unboundStorageBuffer.GetMappedData()) {
+                Memset(mapped, 0, static_cast<SizeT>(kUnboundStorageDescriptorBytes));
+            }
+        }
+        return m_unboundStorageBuffer.GetSlice();
+    }
+
+    BufferSlice VkBufferManager::AcquireUnboundTexelBufferDescriptor() {
+        if (!m_unboundTexelBuffer.IsValid()) {
+            if (m_initInfo.allocator == nullptr) {
+                return {};
+            }
+            // A SECOND placeholder rather than more usage bits on the storage-block one. The two
+            // are independent failure domains: a device that refuses this allocation must not
+            // take the storage-block placeholder - and with it the fix this one is a sibling of -
+            // down with it. Host-visible and zero-filled for the same reason as that one: this is
+            // reached from descriptor resolution, inside an already-open recording, which must
+            // not start a copy of its own.
+            const Bool created = m_unboundTexelBuffer.Create({
+                .allocator = m_initInfo.allocator,
+                .size = kUnboundTexelBufferDescriptorBytes,
+                .usage = VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT |
+                         VK_BUFFER_USAGE_TRANSFER_DST_BIT,
+                .memoryUsage = VMA_MEMORY_USAGE_AUTO,
+                .allocationFlags = VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT |
+                                   VMA_ALLOCATION_CREATE_MAPPED_BIT,
+                .requiredFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
+            });
+            if (!created) {
+                MGLOG_E_ONCE("VkBufferManager::AcquireUnboundTexelBufferDescriptor: placeholder creation failed");
+                m_unboundTexelBuffer.Destroy();
+                return {};
+            }
+            if (void* mapped = m_unboundTexelBuffer.GetMappedData()) {
+                Memset(mapped, 0, static_cast<SizeT>(kUnboundTexelBufferDescriptorBytes));
+            }
+        }
+        return m_unboundTexelBuffer.GetSlice();
+    }
+
     VkBufferUsageFlags VkBufferManager::GetVkBufferUsage(BufferKind kind) {
         switch (kind) {
         case BufferKind::Vertex:
@@ -595,7 +793,13 @@ 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

@@ -31,6 +31,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VmaMemoryUsage transientMemoryUsage = VMA_MEMORY_USAGE_AUTO;
         VmaAllocationCreateFlags transientAllocationFlags = VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
         Bool transientPersistentMapping = false;
+        // VK_EXT_transform_feedback is enabled: persistent-map storage additionally
+        // carries the transform feedback usage so capture targets can bind directly.
+        Bool transformFeedbackUsageEnabled = false;
     };
 
     // The DirectVulkan storage behind one frontend buffer (pipe_resource analogue).
@@ -54,11 +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,
@@ -77,6 +102,12 @@ 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. Only valid when
+        // the caller has proven every queue submission complete; used by the present-less
+        // frame-boundary drain. Deliberately does NOT touch the transient arena's parked
+        // 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();
         // A frame slot's submission fence has been waited: every serial up to
@@ -89,6 +120,27 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Bool UploadTransient(BufferKind kind, Uint32 frameIndex, const void* data, VkDeviceSize size,
                              VkDeviceSize alignment, BufferSlice& outSlice);
 
+        // The descriptor a shader storage block gets when the program declares it and the
+        // application bound no buffer at its GL binding point. GL 4.6 core 7.8 makes that a
+        // legal state - the block simply has no store, so reads are undefined and writes go
+        // nowhere - whereas Vulkan has no such thing as an unwritten descriptor, so something
+        // real has to sit in the set or the whole draw/dispatch is lost. One zero-filled
+        // buffer, created once and shared by every unbound binding: bindings that are only
+        // declared (the case this exists for) never touch it, and one that is actually read
+        // sees zeros, which is inside GL's "undefined". robustBufferAccess bounds anything
+        // that indexes past it.
+        BufferSlice AcquireUnboundStorageDescriptor();
+
+        // The store a texel-buffer descriptor - `samplerBuffer` or `imageBuffer` - gets when the
+        // unit the program's uniform names has no buffer texture on it, or the buffer texture on
+        // it has no GL buffer attached. Both are legal GL states that make a fetch return
+        // undefined values (GL 4.6 core 8.9: a buffer texture with no attached buffer object is
+        // incomplete, and sampling an incomplete texture is undefined - not a lost draw), and both
+        // used to take the whole draw or dispatch with them. The VIEW over this - one per format,
+        // and the descriptor is a VkBufferView, not a buffer - is built by
+        // UniformManager::AcquireUnboundTexelBufferView.
+        BufferSlice AcquireUnboundTexelBufferDescriptor();
+
         // Draw-time acquire for resident (device-storage) buffers: ensures the
         // resource exists and is fully uploaded, marks it used this frame.
         Bool AcquireResidentSlice(BufferKind kind, const SharedPtr<MG_State::GLState::BufferObject>& bufferObject,
@@ -113,6 +165,11 @@ 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.
@@ -139,15 +196,28 @@ 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;
+        // See AcquireUnboundStorageDescriptor. Lazily created, never re-created, torn down
+        // with the manager.
+        VkBufferObject m_unboundStorageBuffer;
+        // See AcquireUnboundTexelBufferDescriptor. Same lifetime rules.
+        VkBufferObject m_unboundTexelBuffer;
         IBufferCopyCommandProvider* m_copyProvider = nullptr;
         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

@@ -8,15 +8,79 @@
 
 #include "VkClearManager.h"
 
+// For the shared ResolveAttachmentLayerCount (and the ToVulkanLevelExtent it is built on): the
+// clear key's layer span has to be the same one the render pass builds its attachment view from.
+#include "VkTextureManager.h"
+
+#include "MG_State/GLState/Core.h"
 #include "MG_Util/Converters/MGToStr/FramebufferEnumConverter.h"
 #include "MG_Util/Converters/MGToStr/TextureEnumConverter.h"
 
+#include <algorithm>
+#include <cmath>
+
 namespace MobileGL::MG_Backend::DirectVulkan {
     static Bool IsCubeMapFaceUploadTarget(TextureUploadTarget target) {
         return target >= TextureUploadTarget::CubeMapPositiveX &&
                target <= TextureUploadTarget::CubeMapNegativeZ;
     }
 
+    VkClearColorValue MakeVkClearColorValue(const ClearAttachmentPayload& payload, Bool formatLacksAlpha) {
+        VkClearColorValue clearValue{};
+        switch (payload.colorEncoding) {
+        case ClearColorEncoding::Int:
+            clearValue.int32[0] = payload.colorInt.x();
+            clearValue.int32[1] = payload.colorInt.y();
+            clearValue.int32[2] = payload.colorInt.z();
+            clearValue.int32[3] = formatLacksAlpha ? 1 : payload.colorInt.w();
+            break;
+        case ClearColorEncoding::Uint:
+            clearValue.uint32[0] = payload.colorUint.x();
+            clearValue.uint32[1] = payload.colorUint.y();
+            clearValue.uint32[2] = payload.colorUint.z();
+            clearValue.uint32[3] = formatLacksAlpha ? 1u : payload.colorUint.w();
+            break;
+        case ClearColorEncoding::Float:
+            clearValue.float32[0] = payload.color.x();
+            clearValue.float32[1] = payload.color.y();
+            clearValue.float32[2] = payload.color.z();
+            clearValue.float32[3] = formatLacksAlpha ? 1.0f : payload.color.w();
+            break;
+        }
+        return clearValue;
+    }
+
+    void PreCompensateSrgbClearColor(ClearAttachmentPayload& payload, VkFormat destinationFormat) {
+        if (payload.colorEncoding != ClearColorEncoding::Float) return;
+        // With GL_FRAMEBUFFER_SRGB enabled GL performs the encoding itself, so the driver doing it
+        // is exactly right and there is nothing to undo.
+        if (MG_State::pGLContext->IsCapabilityEnabled(MobileGL::CapabilityInput::FramebufferSrgb)) return;
+        if (ResolveSrgbAttachmentWriteFormat(destinationFormat, false) == destinationFormat) return;
+
+        // sRGB -> linear (GL 4.6 core 8.24), applied to the colour channels only: alpha is stored
+        // linearly in an sRGB format and must pass through untouched.
+        const auto toLinear = [](Float encoded) {
+            const Float value = std::clamp(encoded, 0.0f, 1.0f);
+            return value <= 0.04045f ? value / 12.92f : std::pow((value + 0.055f) / 1.055f, 2.4f);
+        };
+        payload.color = FloatVec4(toLinear(payload.color.x()), toLinear(payload.color.y()),
+                                  toLinear(payload.color.z()), payload.color.w());
+    }
+
+    void ForceOpaqueClearAlpha(ClearAttachmentPayload& payload) {
+        switch (payload.colorEncoding) {
+        case ClearColorEncoding::Int:
+            payload.colorInt = IntVec4(payload.colorInt.x(), payload.colorInt.y(), payload.colorInt.z(), 1);
+            break;
+        case ClearColorEncoding::Uint:
+            payload.colorUint = UintVec4(payload.colorUint.x(), payload.colorUint.y(), payload.colorUint.z(), 1u);
+            break;
+        case ClearColorEncoding::Float:
+            payload.color = FloatVec4(payload.color.x(), payload.color.y(), payload.color.z(), 1.0f);
+            break;
+        }
+    }
+
     static Bool PendingClearMatchesTextureIdentity(const PendingClearKey& key, const TextureIdentity& identity) {
         return key.texture == identity.texture && key.textureLifetimeId == identity.lifetimeId;
     }
@@ -40,13 +104,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         return ResolveAttachmentBaseArrayLayer(uploadTarget);
     }
 
-    static Uint32 ResolveAttachmentLayerCount(
-        const MG_State::GLState::FramebufferAttachmentObject& attachment) {
-        if (attachment.IsLayered()) {
-            return static_cast<Uint32>(std::max(attachment.GetSize().z(), 1));
-        }
-        return 1u;
-    }
+    // ResolveAttachmentLayerCount used to be duplicated here, reading attachment.GetSize().z()
+    // raw - no ToVulkanLevelExtent remap for a 1D array, no six-faces arm for a cube map. That is
+    // not a cosmetic difference: the count below is not key-only, it is written straight into
+    // VkImageSubresourceRange::layerCount by MaterializePendingClearForTexture, which then POPS
+    // the entry - so a layered cube map's glClear reached one face and the other five were lost
+    // for good, while the very same queued clear cleared all six through the render pass's
+    // LOAD_OP_CLEAR. The helper now lives once, in VkTextureManager.h beside ToVulkanLevelExtent.
 
     static const MG_State::GLState::FramebufferAttachmentObject* GetClearableAttachment(
         const MG_State::GLState::FramebufferObject& drawFbo, FramebufferAttachmentType attachmentType) {
@@ -62,8 +126,30 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         return &attachment;
     }
 
-    PendingClearKey VkClearManager::MakePendingClearKey(MG_State::GLState::ITextureObject* texture, Uint32 mipLevel,
+    // The texture a pending clear is actually ABOUT. A clear issued through a GL texture view
+    // (ARB_texture_view) targets the storage it views, so it must queue against - and be found
+    // by - the storage texture; keying it on the view instead left the clear invisible to every
+    // materialisation done through the parent's name (and vice versa), so the image stayed in
+    // VK_IMAGE_LAYOUT_UNDEFINED and the readback was dropped as unreadable.
+    static MG_State::GLState::ITextureObject* ClearStorageTextureOf(MG_State::GLState::ITextureObject* texture) {
+        if (texture == nullptr) {
+            return nullptr;
+        }
+        const auto& storageOwner = texture->GetViewStorageOwner();
+        return storageOwner ? storageOwner.get() : texture;
+    }
+
+    PendingClearKey VkClearManager::MakePendingClearKey(MG_State::GLState::ITextureObject* rawTexture, Uint32 mipLevel,
                                                         Uint32 baseArrayLayer, Uint32 layerCount) {
+        MG_State::GLState::ITextureObject* texture = ClearStorageTextureOf(rawTexture);
+        if (rawTexture != nullptr && texture != rawTexture) {
+            // The caller named a level and a layer of the VIEW; the key describes the STORAGE, so
+            // both have to be shifted into its numbering (GL 4.6 core 8.18). Without this a clear
+            // of a view's level 0 would collide with a clear of the storage's level 0 even when
+            // the view opened onto level 1.
+            mipLevel += static_cast<Uint32>(rawTexture->GetViewMinLevel());
+            baseArrayLayer += static_cast<Uint32>(rawTexture->GetViewMinLayer());
+        }
         return PendingClearKey {
             .texture = texture,
             .textureLifetimeId = texture ? texture->GetLifetimeId() : 0,
@@ -93,9 +179,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         const std::lock_guard<std::mutex> lock(m_mutex);
         m_pendingClears.clear();
         m_aliveObjects.clear();
+        m_pendingCount.store(static_cast<Uint32>(m_pendingClears.size()), std::memory_order_relaxed);
     }
 
     TextureIdentity VkClearManager::MakeTextureIdentity(MG_State::GLState::ITextureObject* texture) {
+        // Same rule as VkTextureManager::MakeTextureIdentity: a GL texture view is identified by
+        // the storage it views. A clear posted against a view and one posted against its parent
+        // target the same image, so they have to coalesce rather than queue independently.
+        if (texture != nullptr) {
+            const auto& storageOwner = texture->GetViewStorageOwner();
+            if (storageOwner) {
+                texture = storageOwner.get();
+            }
+        }
         return TextureIdentity {
             .texture = texture,
             .lifetimeId = texture ? texture->GetLifetimeId() : 0,
@@ -105,7 +201,15 @@ 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;
@@ -127,6 +231,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             m_pendingClears.erase(key);
         }
         m_aliveObjects.erase(identity);
+        m_pendingCount.store(static_cast<Uint32>(m_pendingClears.size()), std::memory_order_relaxed);
     }
 
     Bool VkClearManager::LockTextureIdentityLocked(const TextureIdentity& identity,
@@ -216,11 +321,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return;
         }
 
-        const PendingClearKey key = MakePendingClearKey(texture.get());
+        const auto& storageOwner = texture->GetViewStorageOwner();
+        const SharedPtr<MG_State::GLState::ITextureObject>& storageTexture = storageOwner ? storageOwner : texture;
+        const PendingClearKey key = MakePendingClearKey(storageTexture.get());
         const std::lock_guard<std::mutex> lock(m_mutex);
-        m_aliveObjects[MakeTextureIdentity(texture.get())] = texture;
+        m_aliveObjects[MakeTextureIdentity(storageTexture.get())] = storageTexture;
         auto& pending = m_pendingClears[key];
         MergeClearPayload(pending, clearPayload);
+        m_pendingCount.store(static_cast<Uint32>(m_pendingClears.size()), std::memory_order_relaxed);
     }
 
     void VkClearManager::QueueClear(const ClearAttachmentPayload& clearPayload,
@@ -234,10 +342,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
 
         const PendingClearKey key = MakePendingClearKey(attachment);
+        // The alive entry must hold the STORAGE object, because the key names it:
+        // LockTextureIdentityLocked cross-checks the two, and registering a view here under its
+        // storage's identity made every lookup of this clear fail that check and silently report
+        // "nothing pending" - which is how a clear issued through a view's framebuffer vanished.
+        const auto& storageOwner = texture->GetViewStorageOwner();
+        const SharedPtr<MG_State::GLState::ITextureObject>& storageTexture = storageOwner ? storageOwner : texture;
         const std::lock_guard<std::mutex> lock(m_mutex);
-        m_aliveObjects[MakeTextureIdentity(texture.get())] = texture;
+        m_aliveObjects[MakeTextureIdentity(storageTexture.get())] = storageTexture;
         auto& pending = m_pendingClears[key];
         MergeClearPayload(pending, clearPayload);
+        m_pendingCount.store(static_cast<Uint32>(m_pendingClears.size()), std::memory_order_relaxed);
     }
 
     Bool VkClearManager::HasPendingClear(MG_State::GLState::ITextureObject* texture) {
@@ -245,6 +360,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return false;
         }
 
+        if (m_pendingCount.load(std::memory_order_relaxed) == 0) {
+            return false;  // per-draw hot path: nothing pending anywhere
+        }
+
+        texture = ClearStorageTextureOf(texture);
         const Uint64 lifetimeId = texture->GetLifetimeId();
         const std::lock_guard<std::mutex> lock(m_mutex);
         for (auto it = m_pendingClears.begin(); it != m_pendingClears.end(); ++it) {
@@ -260,6 +380,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (key.texture == nullptr) {
             return false;
         }
+        if (m_pendingCount.load(std::memory_order_relaxed) == 0) {
+            return false;  // per-draw hot path: nothing pending anywhere
+        }
 
         const std::lock_guard<std::mutex> lock(m_mutex);
         if (m_pendingClears.find(key) == m_pendingClears.end()) {
@@ -287,6 +410,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (key.texture == nullptr) {
             return false;
         }
+        if (m_pendingCount.load(std::memory_order_relaxed) == 0) {
+            return false;  // per-draw hot path: nothing pending anywhere
+        }
 
         const std::lock_guard<std::mutex> lock(m_mutex);
         if (!LockTextureLocked(key, outTexture)) {
@@ -325,7 +451,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (texture == nullptr) {
             return false;
         }
+        if (m_pendingCount.load(std::memory_order_relaxed) == 0) {
+            return false;  // per-draw hot path: nothing pending anywhere
+        }
 
+        texture = ClearStorageTextureOf(texture);
         const Uint64 lifetimeId = texture->GetLifetimeId();
         const std::lock_guard<std::mutex> lock(m_mutex);
         SharedPtr<MG_State::GLState::ITextureObject> liveTexture;
@@ -345,6 +475,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return;
         }
 
+        if (m_pendingCount.load(std::memory_order_relaxed) == 0) {
+            return;  // per-draw hot path: nothing pending anywhere
+        }
         const TextureIdentity identity = MakeTextureIdentity(texture);
         MGLOG_D("%s: Pop all pending clears for texture %d", __func__, texture->GetExternalIndex());
         const std::lock_guard<std::mutex> lock(m_mutex);
@@ -361,6 +494,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             auto it = m_pendingClears.find(key);
             if (it != m_pendingClears.end()) {
                 m_pendingClears.erase(it);
+            m_pendingCount.store(static_cast<Uint32>(m_pendingClears.size()), std::memory_order_relaxed);
             }
         }
 

MobileGL/MG_Backend/DirectVulkan/Renderer/VkClearManager.h

@@ -14,6 +14,7 @@
 #include "MG_Util/Math/VectorTypes.h"
 
 #include <Includes.h>
+#include <atomic>
 #include <unordered_map>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
@@ -23,13 +24,41 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Uint32 stencil{};
     };
 
+    // A colour clear reaches us from one of glClear/ClearBufferfv, ClearBufferiv or
+    // ClearBufferuiv, and Vulkan reads VkClearColorValue's union according to the destination
+    // image's format rather than converting between the members - a float written where an
+    // integer format is expected is reinterpreted bit for bit, not rounded. Remember which entry
+    // point supplied the value so the member written when the clear is materialized matches.
+    enum class ClearColorEncoding : Uint8 { Float, Int, Uint };
+
     struct ClearAttachmentPayload {
         GLbitfield mask = 0;
         FloatVec4 color = FloatVec4(0.0f, 0.0f, 0.0f, 0.0f);
+        ClearColorEncoding colorEncoding = ClearColorEncoding::Float;
+        IntVec4 colorInt = IntVec4(0, 0, 0, 0);
+        UintVec4 colorUint = UintVec4(0u, 0u, 0u, 0u);
         Float depth = 1.0f;
         Uint32 stencil = 0;
     };
 
+    // Builds the clear value for `payload` in the union member its encoding calls for.
+    // `formatLacksAlpha` applies GL's rule that a format without an alpha channel reads as one,
+    // expressed in whichever type matches (GL 4.6 core 15.2.3).
+    VkClearColorValue MakeVkClearColorValue(const ClearAttachmentPayload& payload, Bool formatLacksAlpha);
+
+    // Applies that same rule in place, for the paths that have to bake it into the payload before
+    // the destination is known.
+    void ForceOpaqueClearAlpha(ClearAttachmentPayload& payload);
+
+    // vkCmdClearColorImage names the image, so the driver applies the destination format's transfer
+    // function to whatever value it is handed. Every other write path in this backend goes through
+    // the UNORM twin view while GL_FRAMEBUFFER_SRGB is off (ResolveSrgbAttachmentWriteFormat) and
+    // therefore stores the raw value GL asked for. Rewrites `payload` to the linear colour whose
+    // encoding is that raw value, so a direct image clear of an sRGB destination agrees with them.
+    // A no-op for every other format, for integer clear encodings, and when GL is doing the
+    // encoding itself.
+    void PreCompensateSrgbClearColor(ClearAttachmentPayload& payload, VkFormat destinationFormat);
+
     struct PendingClearKey {
         MG_State::GLState::ITextureObject* texture = nullptr;
         Uint64 textureLifetimeId = 0;
@@ -85,7 +114,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     class VkClearManager {
     public:
         static PendingClearKey MakePendingClearKey(const MG_State::GLState::FramebufferAttachmentObject& attachment);
-        static PendingClearKey MakePendingClearKey(MG_State::GLState::ITextureObject* texture, Uint32 mipLevel = 0,
+        // Resolves a GL texture view to the storage it views before keying; see the definition.
+        static PendingClearKey MakePendingClearKey(MG_State::GLState::ITextureObject* rawTexture, Uint32 mipLevel = 0,
                                                    Uint32 baseArrayLayer = 0, Uint32 layerCount = 1);
 
         Bool Initialize();
@@ -120,7 +150,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                SharedPtr<MG_State::GLState::ITextureObject>& outTexture);
 
         Uint8 m_gcCounter = 0;
+    public:
+        // Lock-free probe for the consecutive-draw fast path: any pending clear
+        // forces the full SetupDraw path (which materializes/consumes it).
+        Bool HasAnyPendingClears() const { return m_pendingCount.load(std::memory_order_relaxed) != 0; }
+
+    private:
         mutable std::mutex m_mutex;
+        // Lock-free mirror of m_pendingClears.size(), maintained under m_mutex
+        // by every mutation. The per-draw probes (HasPendingClear/GetPending*)
+        // read it before taking the lock: during draw batches the pending set
+        // is almost always empty, so this turns several locked map probes per
+        // draw into one relaxed load.
+        std::atomic<Uint32> m_pendingCount{0};
         std::unordered_map<PendingClearKey, ClearAttachmentPayload, PendingClearKeyHash> m_pendingClears;
         std::unordered_map<TextureIdentity, WeakPtr<MG_State::GLState::ITextureObject>, TextureIdentityHash> m_aliveObjects;
     };

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 {
@@ -42,6 +43,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     struct TrackedAttachmentLayoutInfo {
         TrackedAttachmentTarget target = TrackedAttachmentTarget::Texture;
         WeakPtr<MG_State::GLState::ITextureObject> texture;
+        // Identity-compare shortcut for the per-draw "does the active pass use
+        // this sampled texture" probe: comparing this against a LIVE texture's
+        // address needs no weak_ptr::lock (two refcount atomics per probe).
+        // May dangle once the texture dies - compare only, never dereference.
+        MG_State::GLState::ITextureObject* textureRaw = nullptr;
         WeakPtr<MG_State::GLState::RenderbufferObject> renderbuffer;
         Uint32 textureMipLevel = 0;
         Uint32 swapchainImageIndex = 0;
@@ -95,6 +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,
@@ -157,19 +199,68 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     class VkRenderPassManager {
     public:
         using HashType = Uint64;
+
+        // Notified once per OnPresent sweep with every aged-out entry's VkRenderPass
+        // value: pipelines are hashed on the raw handle, and once destroyed the value
+        // may be recycled for an incompatible pass, so dependent caches must purge
+        // everything keyed on them before any new pass can be created (the sweep and
+        // the notification run back-to-back with no creation in between; observers
+        // compare the values, never dereference them). Batched so a mass-idle cohort
+        // (shader-pack switch, dimension exit) costs the observer one pipeline-cache
+        // scan, not one per dying pass. The wholesale paths
+        // (Shutdown/RecreateSwapchain) do not notify - their callers already drop
+        // every pipeline outright.
+        class IEvictionObserver {
+        public:
+            virtual ~IEvictionObserver() = default;
+            virtual void OnRenderPassesDestroyed(const Vector<VkRenderPass>& renderPasses) = 0;
+        };
+
         VkRenderPassManager(VkDevice device,
             VkPhysicalDevice physicalDevice, VmaAllocator allocator, const VulkanRendererConfig& config,
             VkClearManager& clearManager, VkTextureManager& textureManager, SwapchainObject& swapchainObject);
         ~VkRenderPassManager();
 
+        // Observer may be null (no notifications). Not owned.
+        void SetEvictionObserver(IEvictionObserver* observer) { m_evictionObserver = observer; }
+
         Bool Initialize();
         void Shutdown();
 
         HashType ComputeHash(
             const MG_State::GLState::FramebufferObject& fbo,
             Uint32 swapchainImageIndex,
-            Bool includePendingClear = true);
-        RenderPassEntry& GetOrCreateRenderPass(const MG_State::GLState::FramebufferObject& fbo, Uint32 swapchainImageIndex);
+            Bool includePendingClear = true,
+            Bool includeDefaultFboDepthStencil = true);
+        // drawUsesDepthStencil: whether the operation about to run inside the pass
+        // reads or writes the depth/stencil buffer (depth test or stencil test
+        // enabled, or a depth/stencil clear). Only consulted for the DEFAULT
+        // framebuffer: EGL undefines its ancillary buffers at every swap, so a
+        // default-FBO pass whose draws provably never touch depth/stencil is
+        // created WITHOUT the depth attachment - on a tiler that skips the whole
+        // depth tile load AND store. The flavor only escalates: once a pass with
+        // depth is active, later depth-less draws keep using it, and a depth-using
+        // draw against a depth-less active pass resolves to a new (incompatible)
+        // entry, which the caller's compatibility check turns into a pass split;
+        // the new pass's depth loads DONT_CARE (content was undefined all along).
+        //
+        // Returns NULLPTR when this framebuffer cannot be represented as a Vulkan render pass at
+        // all - a texture the texture manager declined to back (an unsupported format or sample
+        // count), or an attachment view it cannot construct (a layer span the image has no room
+        // for, a 3D image whose format was refused 2D-array compatibility). This used to be
+        // unrepresentable: the function returned a reference, so the only thing the two fallible
+        // calls it builds on could do was trip a MOBILEGL_ASSERT - which is compiled out of every
+        // INFO build - and then dereference the null resource, or hand VK_NULL_HANDLE to
+        // vkCreateFramebuffer. That took the whole process down (51 lost CTS records over 21
+        // bodies, one runner restart each) where a declined draw is merely a wrong picture.
+        //
+        // EVERY caller must handle nullptr by dropping the operation, exactly as the draw path
+        // already drops a draw whose sampler descriptor could not be resolved
+        // (UniformManager::BindProgramUniformBuffers). The failure paths log MGLOG_E_ONCE
+        // themselves, so a caller needs no message of its own.
+        [[nodiscard]] RenderPassEntry* GetOrCreateRenderPass(const MG_State::GLState::FramebufferObject& fbo,
+                                                             Uint32 swapchainImageIndex,
+                                                             Bool drawUsesDepthStencil = true);
         void QueueRenderbufferClear(GLbitfield mask, const ClearFramebufferPayload& clearPayload,
                                     const MG_State::GLState::FramebufferObject& drawFbo);
         void QueueRenderbufferClear(const ClearAttachmentPayload& clearPayload,
@@ -192,12 +283,20 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         UnorderedMap<Uint64, RenderPassEntry> m_renderPasses;
         // Monotonic frame counter (bumped in OnPresent) for render-pass cache aging.
         Uint64 m_frameCounter = 0;
+        IEvictionObserver* m_evictionObserver = nullptr;
 
         // Bumped whenever a renderbuffer VkImage is (re)created; together with the texture
         // manager's image epoch this invalidates the render-pass fast path on any attachment
         // image recreation.
         Uint64 m_renderbufferImageEpoch = 1;
 
+    public:
+        // Bumped whenever a renderbuffer backing is (re)created; consecutive-draw
+        // snapshots include it so an attachment respecify forces a re-resolve.
+        Uint64 GetRenderbufferImageEpoch() const { return m_renderbufferImageEpoch; }
+
+    private:
+
         // Per-draw fast-path memo for GetOrCreateRenderPass (dirty-flag state tracking): when the
         // framebuffer state is provably unchanged since the last resolution, the active render pass
         // is reused WITHOUT recomputing the expensive per-draw hash. Invalidated by FBO switch /
@@ -205,17 +304,36 @@ 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;
         Uint64 m_rpFastRbEpoch = 0;
         Uint64 m_rpFastRenderPassHash = 0;
+        // Whether the memoized entry carries a depth/stencil attachment; a
+        // default-FBO resolution whose effective depth request differs must
+        // miss the memo (the depth-less/depth-full flavors hash differently).
+        Bool m_rpFastHadDepthStencil = false;
 
+    public:
         struct RenderbufferResource {
+            // deadSinceFrame sentinel: the owning weak reference has not been observed
+            // expired. Dead resources age past every in-flight frame before Destroy
+            // (see CollectRenderbufferGarbage); the GPU may still reference the image
+            // for frames-in-flight frames after the GL object dies.
+            static constexpr Uint64 kNeverObservedDead = UINT64_MAX;
+
             WeakPtr<MG_State::GLState::RenderbufferObject> renderbuffer;
             VkImage image = VK_NULL_HANDLE;
             VmaAllocation allocation = nullptr;
             VkImageView view = VK_NULL_HANDLE;
+            // UNORM reinterpretation of an sRGB image, used as the attachment view while
+            // GL_FRAMEBUFFER_SRGB is disabled (raw writes). Null for non-sRGB formats.
+            VkImageView unormTwinView = VK_NULL_HANDLE;
             VkImageLayout layout = VK_IMAGE_LAYOUT_UNDEFINED;
             VkFormat format = VK_FORMAT_UNDEFINED;
             VkImageAspectFlags aspect = VK_IMAGE_ASPECT_NONE;
@@ -223,25 +341,75 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VkSampleCountFlagBits sampleCount = VK_SAMPLE_COUNT_1_BIT;
             TextureInternalFormat internalFormat = TextureInternalFormat::Unknown;
             Int samples = 0;
+            // m_frameCounter value at which the weak reference was first seen expired.
+            Uint64 deadSinceFrame = kNeverObservedDead;
 
             void Destroy(VkDevice device, VmaAllocator allocator);
         };
 
+        // Public so the renderer's blit/copy/readback bindings can source renderbuffer
+        // attachments the same way texture attachments go through the texture manager.
+        RenderbufferResource* GetOrCreateRenderbufferResource(
+            const SharedPtr<MG_State::GLState::RenderbufferObject>& renderbuffer);
+        Bool GetPendingRenderbufferClear(MG_State::GLState::RenderbufferObject* renderbuffer,
+                                         ClearAttachmentPayload& outPayload) const;
+
+    private:
         struct PendingRenderbufferClear {
             WeakPtr<MG_State::GLState::RenderbufferObject> renderbuffer;
             ClearAttachmentPayload payload{};
         };
 
-        UnorderedMap<MG_State::GLState::RenderbufferObject*, RenderbufferResource> m_renderbufferResources;
+        // A superseded renderbuffer backing (glRenderbufferStorage respecify) parked
+        // until enough frame boundaries have passed that no in-flight command buffer
+        // can still reference it; destroyed in OnPresent (see RetireAgeFrames).
+        struct DeferredRenderbufferRelease {
+            VkImage image = VK_NULL_HANDLE;
+            VmaAllocation allocation = nullptr;
+            VkImageView view = VK_NULL_HANDLE;
+            VkImageView unormTwinView = VK_NULL_HANDLE;
+            Uint64 deferredAtFrame = 0;
+        };
+
+        // 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
+        // do not repeat vkGetPhysicalDeviceImageFormatProperties.
+        UnorderedMap<VkFormat, VkSampleCountFlags> m_attachmentSampleCountsByFormat;
 
-        RenderbufferResource* GetOrCreateRenderbufferResource(
-            const SharedPtr<MG_State::GLState::RenderbufferObject>& renderbuffer);
-        Bool GetPendingRenderbufferClear(MG_State::GLState::RenderbufferObject* renderbuffer,
-                                         ClearAttachmentPayload& outPayload) const;
         Bool HasPendingRenderbufferClear(
             const MG_State::GLState::FramebufferAttachmentObject& attachment) const;
         void CollectRenderbufferGarbage();
+        // Frame-boundary margin after which a resource last referenced by a retired
+        // GL object (or superseded backing) is provably past every in-flight frame.
+        Uint64 RetireAgeFrames() const;
+        void DeferRenderbufferBackingRelease(RenderbufferResource& resource);
+        void CollectDeferredRenderbufferReleases(Bool destroyAll);
 
         static inline XXH64_state_t* m_hashState = XXH64_createState();
         static inline ActiveRenderPassInfo s_activeRenderPass{};

MobileGL/MG_Backend/DirectVulkan/Renderer/VkSamplerManager.cpp

@@ -21,6 +21,156 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                    sampler.GetWrapR() == SamplerWrapMode::ClampToBorder;
         }
 
+        // The numeric domain the texture is SAMPLED in. Vulkan splits VkBorderColor into a float
+        // family and an integer family and requires the sampler's choice to match the image view's
+        // format (a float border on an integer view, or the reverse, is undefined) - so the domain
+        // comes from the TEXTURE, while the value comes from whichever GL entry point wrote it.
+        enum class BorderColorDomain {
+            Float,
+            SignedInteger,
+            UnsignedInteger
+        };
+
+        BorderColorDomain ResolveBorderColorDomain(TextureInternalFormat format) {
+            switch (format) {
+            case TextureInternalFormat::R8I:
+            case TextureInternalFormat::R16I:
+            case TextureInternalFormat::R32I:
+            case TextureInternalFormat::RG8I:
+            case TextureInternalFormat::RG16I:
+            case TextureInternalFormat::RG32I:
+            case TextureInternalFormat::RGB8I:
+            case TextureInternalFormat::RGB16I:
+            case TextureInternalFormat::RGB32I:
+            case TextureInternalFormat::RGBA8I:
+            case TextureInternalFormat::RGBA16I:
+            case TextureInternalFormat::RGBA32I:
+                return BorderColorDomain::SignedInteger;
+            case TextureInternalFormat::R8UI:
+            case TextureInternalFormat::R16UI:
+            case TextureInternalFormat::R32UI:
+            case TextureInternalFormat::RG8UI:
+            case TextureInternalFormat::RG16UI:
+            case TextureInternalFormat::RG32UI:
+            case TextureInternalFormat::RGB8UI:
+            case TextureInternalFormat::RGB16UI:
+            case TextureInternalFormat::RGB32UI:
+            case TextureInternalFormat::RGBA8UI:
+            case TextureInternalFormat::RGBA16UI:
+            case TextureInternalFormat::RGBA32UI:
+            case TextureInternalFormat::RGB10A2UI:
+                return BorderColorDomain::UnsignedInteger;
+            default:
+                return BorderColorDomain::Float;
+            }
+        }
+
+        Bool IsSignedNormalizedFormat(TextureInternalFormat format) {
+            switch (format) {
+            case TextureInternalFormat::R8Snorm:
+            case TextureInternalFormat::R16Snorm:
+            case TextureInternalFormat::RG8Snorm:
+            case TextureInternalFormat::RG16Snorm:
+            case TextureInternalFormat::RGB8Snorm:
+            case TextureInternalFormat::RGB16Snorm:
+            case TextureInternalFormat::RGBA8Snorm:
+            case TextureInternalFormat::RGBA16Snorm:
+                return true;
+            default:
+                return false;
+            }
+        }
+
+        // GL 4.6 core 8.14.2: "The border values are clamped before they are used, according to the
+        // format in which texture components are stored. For signed and unsigned normalized
+        // fixed-point formats, border values are clamped to [-1,1] and [0,1] respectively. For
+        // floating-point and integer formats, border values are clamped to the representable range of
+        // the format." Every clause of that sentence is a real case here - the clamp is not just the
+        // normalized one.
+        //
+        // Only the 32-bit float formats are genuinely unclamped: every finite float is representable
+        // in them. Half-float has a finite maximum, and the two packed "float" formats are UNSIGNED,
+        // so a negative border on them must come back as 0 rather than as a negative number the
+        // driver delivers verbatim through VK_BORDER_COLOR_FLOAT_CUSTOM_EXT.
+        struct FloatBorderRange {
+            Bool clamped = true;
+            Float minValue = 0.0f;
+            Float maxValue = 1.0f;
+        };
+
+        FloatBorderRange ResolveFloatBorderRange(TextureInternalFormat format, Bool isSignedNormalized) {
+            switch (format) {
+            case TextureInternalFormat::R32F:
+            case TextureInternalFormat::RG32F:
+            case TextureInternalFormat::RGB32F:
+            case TextureInternalFormat::RGBA32F:
+                return {false, 0.0f, 0.0f};
+            case TextureInternalFormat::R16F:
+            case TextureInternalFormat::RG16F:
+            case TextureInternalFormat::RGB16F:
+            case TextureInternalFormat::RGBA16F:
+                return {true, -65504.0f, 65504.0f};
+            // Unsigned packed floats: no sign bit at all. 65024 is the largest 11-bit float; the
+            // 10-bit blue channel tops out lower (64512) and RGB9E5 higher (65408), but the bound
+            // that matters for correctness is the lower one, and a single conservative upper bound
+            // costs nothing a real border colour will ever notice.
+            case TextureInternalFormat::R11FG11FB10F:
+                return {true, 0.0f, 64512.0f};
+            case TextureInternalFormat::RGB9E5:
+                return {true, 0.0f, 65408.0f};
+            default:
+                return {true, isSignedNormalized ? -1.0f : 0.0f, 1.0f};
+            }
+        }
+
+        // Per-component representable range of an integer texture format, as Int64 so that the whole
+        // signed and unsigned 32-bit ranges are expressible in one type and the clamp can be written
+        // once for both domains. Alpha is carried separately because RGB10_A2UI is the one format
+        // whose alpha is narrower than its colour channels.
+        struct IntegerBorderRange {
+            Int64 rgbMin = 0;
+            Int64 rgbMax = 0;
+            Int64 alphaMin = 0;
+            Int64 alphaMax = 0;
+        };
+
+        IntegerBorderRange ResolveIntegerBorderRange(TextureInternalFormat format) {
+            const auto uniform = [](Int64 low, Int64 high) { return IntegerBorderRange{low, high, low, high}; };
+            switch (format) {
+            case TextureInternalFormat::R8I:
+            case TextureInternalFormat::RG8I:
+            case TextureInternalFormat::RGB8I:
+            case TextureInternalFormat::RGBA8I:
+                return uniform(-128, 127);
+            case TextureInternalFormat::R16I:
+            case TextureInternalFormat::RG16I:
+            case TextureInternalFormat::RGB16I:
+            case TextureInternalFormat::RGBA16I:
+                return uniform(-32768, 32767);
+            case TextureInternalFormat::R8UI:
+            case TextureInternalFormat::RG8UI:
+            case TextureInternalFormat::RGB8UI:
+            case TextureInternalFormat::RGBA8UI:
+                return uniform(0, 255);
+            case TextureInternalFormat::R16UI:
+            case TextureInternalFormat::RG16UI:
+            case TextureInternalFormat::RGB16UI:
+            case TextureInternalFormat::RGBA16UI:
+                return uniform(0, 65535);
+            case TextureInternalFormat::R32UI:
+            case TextureInternalFormat::RG32UI:
+            case TextureInternalFormat::RGB32UI:
+            case TextureInternalFormat::RGBA32UI:
+                return uniform(0, 4294967295LL);
+            case TextureInternalFormat::RGB10A2UI:
+                return {0, 1023, 0, 3};
+            default:
+                // The signed 32-bit formats, and anything unexpected: the full int32 range, i.e. a
+                // clamp that cannot alter a value the GL entry points could have carried.
+                return uniform(-2147483648LL, 2147483647LL);
+            }
+        }
+
         Bool IsDepthTextureFormat(TextureInternalFormat format) {
             switch (format) {
             case TextureInternalFormat::DepthComponent:
@@ -51,6 +201,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Float ResolveEffectiveMinLod(const MG_State::GLState::SamplerObject& sampler, Float effectiveMaxLod) {
             return std::min(sampler.GetMinLod(), effectiveMaxLod);
         }
+
+        // A single-level view can only ever deliver the base level, but the LOD clamp must not be
+        // collapsed to exactly 0: both GL and Vulkan pick magFilter over minFilter from the
+        // *clamped* lambda, so maxLod = 0 would make every fragment magnify and quietly retire the
+        // min filter. 0.25 is the value VkSamplerCreateInfo's own note prescribes for emulating
+        // GL's non-mipmapped minification - large enough for lambda to stay positive, small enough
+        // that a NEAREST mip mode still rounds down to level 0. Clamped rather than assigned, so a
+        // texture whose GL_TEXTURE_MAX_LOD really is 0 keeps magnifying as GL says it must.
+        Float ResolveSingleLevelMaxLod(const MG_State::GLState::SamplerObject& sampler, Bool singleLevelView) {
+            const Float maxLod = ResolveEffectiveMaxLod(sampler);
+            return singleLevelView ? std::min(maxLod, 0.25f) : maxLod;
+        }
     } // namespace
 
     Bool VkSamplerManager::Initialize(const InitInfo& initInfo) {
@@ -60,6 +222,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_config = initInfo.config;
         m_samplerAnisotropySupported = initInfo.samplerAnisotropySupported;
         m_maxSamplerAnisotropy = std::max(initInfo.maxSamplerAnisotropy, 1.0f);
+        m_customBorderColorSupported = initInfo.customBorderColorSupported;
+        m_maxCustomBorderColorSamplers = initInfo.maxCustomBorderColorSamplers;
+        m_customBorderColorSamplerCount = 0;
         MOBILEGL_ASSERT(m_device != VK_NULL_HANDLE && m_config != nullptr,
                         "VkSamplerManager::Initialize failed: invalid initialization info");
         return true;
@@ -89,15 +254,49 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         m_device = VK_NULL_HANDLE;
         m_config = nullptr;
+        m_frameBoundaryCounter = 0;
+        m_customBorderColorSupported = false;
+        m_maxCustomBorderColorSamplers = 0;
+        m_customBorderColorSamplerCount = 0;
+    }
+
+    void VkSamplerManager::OnFrameBoundary() {
+        ++m_frameBoundaryCounter;
+
+        // Sweep occasionally; destroy samplers whose last use is far past every
+        // in-flight frame. Destroy and erase must stay atomic, or Shutdown would
+        // double-free the handle; an evicted key that recurs simply re-creates
+        // its sampler on the next miss.
+        constexpr Uint64 kSweepInterval = 256;
+        constexpr Uint64 kRetireAgeBoundaries = 1024;
+        if ((m_frameBoundaryCounter % kSweepInterval) != 0) {
+            return;
+        }
+
+        for (auto it = m_samplers.begin(); it != m_samplers.end();) {
+            auto& entry = it->second;
+            if (m_frameBoundaryCounter - entry.lastUsedFrameBoundary > kRetireAgeBoundaries) {
+                if (m_device != VK_NULL_HANDLE && entry.handle != VK_NULL_HANDLE) {
+                    vkDestroySampler(m_device, entry.handle, nullptr);
+                }
+                if (entry.usesCustomBorderColor && m_customBorderColorSamplerCount > 0) {
+                    --m_customBorderColorSamplerCount;
+                }
+                it = m_samplers.erase(it);
+            } else {
+                ++it;
+            }
+        }
     }
 
     Uint64 VkSamplerManager::BuildSamplerKey(const MG_State::GLState::SamplerObject& sampler,
-                                             const MG_State::GLState::ITextureObject& texture,
-                                             Bool forceNearestFiltering) const {
+                                             Bool forceNearestFiltering, Bool singleLevelView,
+                                             const ResolvedBorderColor& borderColor) const {
         MOBILEGL_ASSERT(m_config != nullptr, "VkSamplerManager::BuildSamplerKey: m_config is null");
         XXHASH_VERIFY(XXH64_reset(m_hashState, m_config->CacheVersion));
 
         XXHASH_VERIFY(XXH64_update(m_hashState, &forceNearestFiltering, sizeof(forceNearestFiltering)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &singleLevelView, sizeof(singleLevelView)));
 
         const auto minFilter = sampler.GetMinFilter();
         XXHASH_VERIFY(XXH64_update(m_hashState, &minFilter, sizeof(minFilter)));
@@ -111,7 +310,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         XXHASH_VERIFY(XXH64_update(m_hashState, &wrapT, sizeof(wrapT)));
         const auto wrapR = sampler.GetWrapR();
         XXHASH_VERIFY(XXH64_update(m_hashState, &wrapR, sizeof(wrapR)));
-        const auto maxLod = ResolveEffectiveMaxLod(sampler);
+        const auto maxLod = ResolveSingleLevelMaxLod(sampler, singleLevelView);
         const auto minLod = ResolveEffectiveMinLod(sampler, maxLod);
         XXHASH_VERIFY(XXH64_update(m_hashState, &minLod, sizeof(minLod)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &maxLod, sizeof(maxLod)));
@@ -124,19 +323,36 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         XXHASH_VERIFY(XXH64_update(m_hashState, &maxAnisotropy, sizeof(maxAnisotropy)));
         const auto compareMode = sampler.GetCompareMode();
         XXHASH_VERIFY(XXH64_update(m_hashState, &compareMode, sizeof(compareMode)));
-        const auto compareFunc = ResolveCompareFunc(sampler, texture);
+        const auto compareFunc = sampler.GetSamplerCompareFunc();
         XXHASH_VERIFY(XXH64_update(m_hashState, &compareFunc, sizeof(compareFunc)));
-        const auto borderColor = ResolveVkBorderColor(sampler, texture);
-        XXHASH_VERIFY(XXH64_update(m_hashState, &borderColor, sizeof(borderColor)));
+        // The resolved enum AND, when it is one of the *_CUSTOM_EXT values, the sixteen bytes of the
+        // colour itself: two samplers that differ only in a custom border colour carry the same enum
+        // and would otherwise collide onto whichever one was created first.
+        XXHASH_VERIFY(XXH64_update(m_hashState, &borderColor.color, sizeof(borderColor.color)));
+        if (borderColor.isCustom) {
+            XXHASH_VERIFY(XXH64_update(m_hashState, &borderColor.customValue, sizeof(borderColor.customValue)));
+        }
         return XXH64_digest(m_hashState);
     }
 
     VkSampler VkSamplerManager::GetOrCreateSampler(const MG_State::GLState::SamplerObject& sampler,
                                                    const MG_State::GLState::ITextureObject& texture,
-                                                   Bool forceNearestFiltering) {
-        const Uint64 key = BuildSamplerKey(sampler, texture, forceNearestFiltering);
+                                                   Bool forceNearestFiltering, Uint32 viewLevelCount) {
+        // A view that exposes a single mip level has no second level to blend with, so GL's
+        // *_MIPMAP_* minification filters degenerate to plain filtering on the base level -
+        // sampling is unchanged by pinning the Vulkan sampler to NEAREST mip mode at LOD 0.
+        // It is not cosmetic: MobileGL backs such a view with a fully allocated mip chain whose
+        // tail is never written, and a LINEAR mip mode lets the texture unit issue the level+1
+        // fetch anyway. On Adreno that fetch lands in uninitialized UBWC pages (or past the
+        // allocation for a genuinely single-level image) and faults the GPU - the same failure
+        // the default-framebuffer blit shader had to work around with an explicit-LOD sample.
+        const Bool singleLevelView = viewLevelCount == 1;
+        // Resolved once and used for both the key and the create-info; see ResolvedBorderColor.
+        const ResolvedBorderColor borderColor = ResolveBorderColor(sampler, texture);
+        const Uint64 key = BuildSamplerKey(sampler, forceNearestFiltering, singleLevelView, borderColor);
         auto it = m_samplers.find(key);
         if (it != m_samplers.end()) {
+            it->second.lastUsedFrameBoundary = m_frameBoundaryCounter;
             return it->second.handle;
         }
 
@@ -144,8 +360,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         samplerInfo.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO;
         samplerInfo.magFilter = forceNearestFiltering ? VK_FILTER_NEAREST : ToVkFilter(sampler.GetMagFilter());
         samplerInfo.minFilter = forceNearestFiltering ? VK_FILTER_NEAREST : ToVkFilter(sampler.GetMinFilter());
-        samplerInfo.mipmapMode = forceNearestFiltering ? VK_SAMPLER_MIPMAP_MODE_NEAREST
-                                                       : ToVkMipmapMode(sampler.GetMipmapMode());
+        samplerInfo.mipmapMode = (forceNearestFiltering || singleLevelView)
+                                     ? VK_SAMPLER_MIPMAP_MODE_NEAREST
+                                     : ToVkMipmapMode(sampler.GetMipmapMode());
         samplerInfo.addressModeU = ToVkAddressMode(sampler.GetWrapS());
         samplerInfo.addressModeV = ToVkAddressMode(sampler.GetWrapT());
         samplerInfo.addressModeW = ToVkAddressMode(sampler.GetWrapR());
@@ -156,12 +373,25 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         samplerInfo.anisotropyEnable = maxAnisotropy > 1.0f ? VK_TRUE : VK_FALSE;
         samplerInfo.maxAnisotropy = maxAnisotropy;
         samplerInfo.compareEnable = sampler.GetCompareMode() == SamplerCompareMode::CompareToTexture ? VK_TRUE : VK_FALSE;
-        samplerInfo.compareOp = ToVkCompareOp(ResolveCompareFunc(sampler, texture));
-        samplerInfo.maxLod = ResolveEffectiveMaxLod(sampler);
+        samplerInfo.compareOp = ToVkCompareOp(sampler.GetSamplerCompareFunc());
+        // Must match BuildSamplerKey's resolution exactly.
+        samplerInfo.maxLod = ResolveSingleLevelMaxLod(sampler, singleLevelView);
         samplerInfo.minLod = ResolveEffectiveMinLod(sampler, samplerInfo.maxLod);
-        samplerInfo.borderColor = ResolveVkBorderColor(sampler, texture);
+        samplerInfo.borderColor = borderColor.color;
         samplerInfo.unnormalizedCoordinates = VK_FALSE;
 
+        // VK_EXT_custom_border_color. `format` stays UNDEFINED, which is legal only because
+        // customBorderColorWithoutFormat was required alongside customBorderColors at device
+        // creation - a GL sampler object has no idea which texture it will be paired with.
+        VkSamplerCustomBorderColorCreateInfoEXT customBorderColorInfo{};
+        if (borderColor.isCustom) {
+            customBorderColorInfo.sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT;
+            customBorderColorInfo.customBorderColor = borderColor.customValue;
+            customBorderColorInfo.format = VK_FORMAT_UNDEFINED;
+            customBorderColorInfo.pNext = samplerInfo.pNext;
+            samplerInfo.pNext = &customBorderColorInfo;
+        }
+
         VkSampler vkSampler = VK_NULL_HANDLE;
         VK_VERIFY(vkCreateSampler(m_device, &samplerInfo, nullptr, &vkSampler), "vkCreateSampler(texture)");
 
@@ -169,6 +399,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         entry.handle = vkSampler;
         entry.externalIndex = sampler.GetExternalIndex();
         entry.version = sampler.GetVersion();
+        entry.lastUsedFrameBoundary = m_frameBoundaryCounter;
+        entry.usesCustomBorderColor = borderColor.isCustom;
+        if (entry.usesCustomBorderColor) {
+            ++m_customBorderColorSamplerCount;
+        }
         m_samplers[key] = entry;
         return vkSampler;
     }
@@ -228,48 +463,148 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
     }
 
-    SamplerCompareFunc VkSamplerManager::ResolveCompareFunc(const MG_State::GLState::SamplerObject& sampler,
-                                                            const MG_State::GLState::ITextureObject& texture) {
-        const auto compareFunc = sampler.GetSamplerCompareFunc();
-        if (sampler.GetCompareMode() == SamplerCompareMode::CompareToTexture &&
-            IsDepthTextureFormat(texture.GetFormat()) && compareFunc == SamplerCompareFunc::Always) {
-            return SamplerCompareFunc::LessEqual;
-        }
-
-        return compareFunc;
-    }
-
-    VkBorderColor VkSamplerManager::ResolveVkBorderColor(const MG_State::GLState::SamplerObject& sampler,
-                                                         const MG_State::GLState::ITextureObject& texture) {
+    VkSamplerManager::ResolvedBorderColor VkSamplerManager::ResolveBorderColor(
+        const MG_State::GLState::SamplerObject& sampler, const MG_State::GLState::ITextureObject& texture) const {
+        ResolvedBorderColor resolved{};
         if (!UsesBorderColor(sampler)) {
-            return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+            return resolved; // FLOAT_TRANSPARENT_BLACK, never sampled
         }
 
-        const auto& borderColor = texture.GetBorderColor();
-        const Bool isDepthTexture = IsDepthTextureFormat(texture.GetFormat());
+        // Border colour is sampler state: a bound sampler object supplies its own, and a texture
+        // with none reaches the very same value through the sampler object it owns.
+        const auto format = texture.GetFormat();
+        const auto domain = ResolveBorderColorDomain(format);
+        const Bool canUseCustom = m_customBorderColorSupported && m_maxCustomBorderColorSamplers > 0 &&
+                                  m_customBorderColorSamplerCount < m_maxCustomBorderColorSamplers;
 
-        if (isDepthTexture) {
+        if (domain != BorderColorDomain::Float) {
+            // An integer image view REQUIRES an integer border colour, whatever the value is - even
+            // (0,0,0,1). The value itself is whichever integer form the application wrote; a float
+            // border on an integer texture is nonsense GL leaves undefined, so the derived integer
+            // representation (a plain cast) is as good an answer as any.
+            //
+            // Clamped to the format's representable range FIRST, per GL 4.6 core 8.14.2, and read
+            // through Int64 so the whole signed and unsigned 32-bit ranges are expressible at once.
+            //
+            // Which representation to start from is the TEXTURE's domain, not the entry-point form
+            // the application used. GL 4.6 core 8.10 stores an "I"-form border colour unmodified with
+            // an integer internal data type and does not define a sign conversion between the two
+            // integer forms, so the stored bits are reinterpreted in the sampled format's own
+            // signedness. Measured, not assumed: a border of -1 written with glTexParameterIiv
+            // against a GL_R8UI texture samples as 255 on the ES driver, i.e. as 0xFFFFFFFF clamped
+            // to the format's maximum - see the IntegerBorderColorScenario case that pins it. Picking
+            // the representation by the FORM instead would answer 0 here, which is a defensible
+            // reading of the same spec text but puts DirectVulkan at odds with DirectGLES - and
+            // DirectGLES cannot deviate, it forwards the value to the driver verbatim. Cross-backend
+            // agreement decides it.
+            const auto range = ResolveIntegerBorderRange(format);
+            const auto& borderColorI = sampler.GetBorderColorI();
+            const auto& borderColorUI = sampler.GetBorderColorUI();
+            const Bool startFromUnsigned = domain == BorderColorDomain::UnsignedInteger;
+            Int64 clamped[4];
+            for (SizeT channel = 0; channel < 4; ++channel) {
+                const Int64 raw = startFromUnsigned ? static_cast<Int64>(borderColorUI[channel])
+                                                    : static_cast<Int64>(borderColorI[channel]);
+                const Int64 low = channel == 3 ? range.alphaMin : range.rgbMin;
+                const Int64 high = channel == 3 ? range.alphaMax : range.rgbMax;
+                clamped[channel] = std::clamp(raw, low, high);
+            }
+
+            // Matched against the CLAMPED value, so a border the format cannot hold still lands on
+            // the palette entry it clamps to rather than missing every one of them.
+            const Bool allZeroRgb = clamped[0] == 0 && clamped[1] == 0 && clamped[2] == 0;
+            if (allZeroRgb && clamped[3] == 0) {
+                resolved.color = VK_BORDER_COLOR_INT_TRANSPARENT_BLACK;
+                return resolved;
+            }
+            if (allZeroRgb && clamped[3] == 1) {
+                resolved.color = VK_BORDER_COLOR_INT_OPAQUE_BLACK;
+                return resolved;
+            }
+            if (clamped[0] == 1 && clamped[1] == 1 && clamped[2] == 1 && clamped[3] == 1) {
+                resolved.color = VK_BORDER_COLOR_INT_OPAQUE_WHITE;
+                return resolved;
+            }
+            if (canUseCustom) {
+                resolved.color = VK_BORDER_COLOR_INT_CUSTOM_EXT;
+                resolved.isCustom = true;
+                for (SizeT channel = 0; channel < 4; ++channel) {
+                    if (domain == BorderColorDomain::UnsignedInteger) {
+                        resolved.customValue.uint32[channel] = static_cast<Uint32>(clamped[channel]);
+                    } else {
+                        resolved.customValue.int32[channel] = static_cast<Int32>(clamped[channel]);
+                    }
+                }
+                return resolved;
+            }
+            // No custom colour available: pick the nearest of the three integer palette entries
+            // rather than always answering transparent black, which is what turned an integer border
+            // of (-1,-1,-1,-1) into 0 and broke the CTS's clamped-texel detection outright.
+            const Bool opaque = clamped[3] != 0;
+            const Bool bright = clamped[0] != 0 || clamped[1] != 0 || clamped[2] != 0;
+            resolved.color = !opaque ? VK_BORDER_COLOR_INT_TRANSPARENT_BLACK
+                                     : (bright ? VK_BORDER_COLOR_INT_OPAQUE_WHITE : VK_BORDER_COLOR_INT_OPAQUE_BLACK);
+            return resolved;
+        }
+
+        // Float domain. GL 4.6 core 8.14.2/8.23: the border colour is interpreted in the texture's
+        // format, so it is clamped to that format's representable range first. Without the clamp the
+        // CTS's border of (255,255,255,255) on a GL_RGBA8 texture matched none of the palette entries
+        // and fell through to transparent black - every border texel sampled 0 where the test wanted
+        // 255. The range is per format class, not just the normalized [0,1] / [-1,1] pair: only the
+        // 32-bit float formats are unclamped.
+        FloatVec4 borderColor = sampler.GetBorderColor();
+        if (const auto range = ResolveFloatBorderRange(format, IsSignedNormalizedFormat(format)); range.clamped) {
+            borderColor = FloatVec4(std::clamp(borderColor.x(), range.minValue, range.maxValue),
+                                    std::clamp(borderColor.y(), range.minValue, range.maxValue),
+                                    std::clamp(borderColor.z(), range.minValue, range.maxValue),
+                                    std::clamp(borderColor.w(), range.minValue, range.maxValue));
+        }
+
+        // A depth texture samples one component, so only x decides - and its alpha reads as 1.
+        if (IsDepthTextureFormat(format)) {
             if (NearlyEqual(borderColor.x(), 1.0f)) {
-                return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
+                resolved.color = VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
+                return resolved;
             }
             if (NearlyEqual(borderColor.x(), 0.0f)) {
-                return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+                resolved.color = VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+                return resolved;
             }
         }
 
         const Bool rgbZero = NearlyEqual(borderColor.x(), 0.0f) && NearlyEqual(borderColor.y(), 0.0f) &&
                              NearlyEqual(borderColor.z(), 0.0f);
         if (rgbZero && NearlyEqual(borderColor.w(), 0.0f)) {
-            return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+            resolved.color = VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+            return resolved;
         }
         if (rgbZero && NearlyEqual(borderColor.w(), 1.0f)) {
-            return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+            resolved.color = VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+            return resolved;
         }
         if (NearlyEqual(borderColor.x(), 1.0f) && NearlyEqual(borderColor.y(), 1.0f) &&
             NearlyEqual(borderColor.z(), 1.0f) && NearlyEqual(borderColor.w(), 1.0f)) {
-            return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
+            resolved.color = VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
+            return resolved;
         }
 
-        return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+        if (canUseCustom) {
+            resolved.color = VK_BORDER_COLOR_FLOAT_CUSTOM_EXT;
+            resolved.isCustom = true;
+            resolved.customValue.float32[0] = borderColor.x();
+            resolved.customValue.float32[1] = borderColor.y();
+            resolved.customValue.float32[2] = borderColor.z();
+            resolved.customValue.float32[3] = borderColor.w();
+            return resolved;
+        }
+
+        // Nearest of the three float palette entries. Transparent black stays the answer for a
+        // transparent border, which is what the old unconditional fallback got right by accident.
+        const Bool opaque = borderColor.w() >= 0.5f;
+        const Bool bright = (borderColor.x() + borderColor.y() + borderColor.z()) >= 1.5f;
+        resolved.color = !opaque ? VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK
+                                 : (bright ? VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE : VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK);
+        return resolved;
     }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkSamplerManager.h

@@ -28,33 +28,72 @@ public:
         Bool samplerAnisotropySupported = false;
         // VkPhysicalDeviceLimits::maxSamplerAnisotropy.
         Float maxSamplerAnisotropy = 1.0f;
+        // VK_EXT_custom_border_color was enabled with BOTH customBorderColors and
+        // customBorderColorWithoutFormat; see VulkanRenderer::m_customBorderColorFeatureEnabled.
+        Bool customBorderColorSupported = false;
+        // VkPhysicalDeviceCustomBorderColorPropertiesEXT::maxCustomBorderColorSamplers. A hard device
+        // limit on how many LIVE samplers may carry a custom border colour, so the cache counts them
+        // and falls back to the snapped predefined value once it is reached.
+        Uint32 maxCustomBorderColorSamplers = 0;
     };
 
     Bool Initialize(const InitInfo& initInfo);
     void Shutdown();
 
+    // viewLevelCount is the mip-level count of the image view this sampler will be paired
+    // with; 0 means "unknown, do not narrow". See GetOrCreateSampler for why it matters.
     VkSampler GetOrCreateSampler(const MG_State::GLState::SamplerObject& sampler,
                                  const MG_State::GLState::ITextureObject& texture,
-                                 Bool forceNearestFiltering = false);
+                                 Bool forceNearestFiltering = false,
+                                 Uint32 viewLevelCount = 0);
+    // Frame boundary hook: ages the sampler cache and destroys samplers not used
+    // for many frames. The key hashes continuous float state (lodBias, LOD clamps,
+    // anisotropy), so an app animating those would otherwise mint an unbounded
+    // stream of never-destroyed VkSamplers and eventually exhaust the device's
+    // maxSamplerAllocationCount. A sampler idle for over a thousand frame
+    // boundaries cannot be referenced by any in-flight command buffer (frames in
+    // flight are single digits), and every descriptor set the GPU consumes is
+    // written that same frame with live handles (the per-binding resolve memo and
+    // descriptor-set reuse are both frame-reset), so destruction here needs no
+    // fence wait. Self-gated: one counter bump and compare except on sweep
+    // boundaries.
+    void OnFrameBoundary();
+
+    // What GL_TEXTURE_BORDER_COLOR resolves to for one (sampler, texture) pair. `color` is always a
+    // legal VkBorderColor; when `isCustom` it is one of the *_CUSTOM_EXT values and `customValue`
+    // carries the actual components in a VkSamplerCustomBorderColorCreateInfoEXT.
+    //
+    // Resolved ONCE per GetOrCreateSampler call and threaded into both the cache key and the
+    // create-info, so the two cannot disagree - the same discipline the resolved anisotropy needs,
+    // and here it also makes the maxCustomBorderColorSamplers fallback deterministic: whether a
+    // custom colour was affordable is decided before the key is built, not twice with a budget
+    // change in between.
+    struct ResolvedBorderColor {
+        VkBorderColor color = VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+        VkClearColorValue customValue{};
+        Bool isCustom = false;
+    };
 
 private:
     struct SamplerCacheEntry {
         VkSampler handle = VK_NULL_HANDLE;
         Uint externalIndex = 0;
         Uint16 version = 0;
+        // Frame boundary of the last cache hit; entries idle past the
+        // OnFrameBoundary retirement age have their VkSampler destroyed.
+        Uint64 lastUsedFrameBoundary = 0;
+        // Counted against maxCustomBorderColorSamplers for as long as this entry lives.
+        Bool usesCustomBorderColor = false;
     };
 
-    Uint64 BuildSamplerKey(const MG_State::GLState::SamplerObject& sampler,
-                           const MG_State::GLState::ITextureObject& texture,
-                           Bool forceNearestFiltering) const;
+    Uint64 BuildSamplerKey(const MG_State::GLState::SamplerObject& sampler, Bool forceNearestFiltering,
+                           Bool singleLevelView, const ResolvedBorderColor& borderColor) const;
     static VkFilter ToVkFilter(SamplerFilterMode mode);
     static VkSamplerMipmapMode ToVkMipmapMode(SamplerMipmapMode mode);
     static VkSamplerAddressMode ToVkAddressMode(SamplerWrapMode mode);
     static VkCompareOp ToVkCompareOp(SamplerCompareFunc func);
-    static SamplerCompareFunc ResolveCompareFunc(const MG_State::GLState::SamplerObject& sampler,
-                                                 const MG_State::GLState::ITextureObject& texture);
-    static VkBorderColor ResolveVkBorderColor(const MG_State::GLState::SamplerObject& sampler,
-                                              const MG_State::GLState::ITextureObject& texture);
+    ResolvedBorderColor ResolveBorderColor(const MG_State::GLState::SamplerObject& sampler,
+                                           const MG_State::GLState::ITextureObject& texture) const;
     // The anisotropy Vulkan will actually apply: 1.0 (i.e. disabled) unless the feature is on and
     // the sampler filters linearly both ways, otherwise the GL request clamped to the device limit.
     // GL happily carries GL_TEXTURE_MAX_ANISOTROPY on a NEAREST sampler (Blaze3D's blocks do exactly
@@ -66,7 +105,15 @@ private:
     const VulkanRendererConfig* m_config = nullptr;
     Bool m_samplerAnisotropySupported = false;
     Float m_maxSamplerAnisotropy = 1.0f;
+    Bool m_customBorderColorSupported = false;
+    Uint32 m_maxCustomBorderColorSamplers = 0;
+    // Live cache entries carrying a custom border colour. Kept in step with the entries themselves
+    // in exactly the three places one can appear or disappear: creation, the OnFrameBoundary sweep,
+    // and Shutdown.
+    Uint32 m_customBorderColorSamplerCount = 0;
     UnorderedMap<Uint64, SamplerCacheEntry> m_samplers;
+    // Monotonic frame-boundary counter (bumped in OnFrameBoundary) for cache aging.
+    Uint64 m_frameBoundaryCounter = 0;
     static inline XXH64_state_t* m_hashState = XXH64_createState();
 };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkTextureManager.h

@@ -10,9 +10,12 @@
 
 #include "../VkIncludes.h"
 #include <Includes.h>
+#include <MG_State/GLState/FramebufferState/FramebufferObject.h>
 #include <MG_State/GLState/TextureState/TextureObject.h>
 #include <vk_mem_alloc.h>
+#include <algorithm>
 #include <unordered_map>
+#include <unordered_set>
 
 namespace MobileGL::MG_State::GLState {
 class ITextureObject;
@@ -21,12 +24,111 @@ class ITextureObject;
 namespace MobileGL::MG_Backend::DirectVulkan {
 enum class SamplerNumericDomain : Uint8;
 
+// What VkFormat a GL internal format is BACKED with, and how a shadow upload has to be reshaped to
+// fit it. This is not the same question as "is there an exact VkFormat for this GL format", which is
+// what ConvertTextureInternalFormatToVkEnum answers: several GL formats have no Vulkan twin at all
+// (RGBA2, RGBA12) and several three-channel ones are deliberately widened to their four-channel twin
+// because Vulkan devices rarely support the 3-channel layouts.
+//
+// SHARED, and it must stay the only answer to that question. A renderbuffer and a texture of the
+// same GL format have to resolve to the SAME VkFormat or every blit, resolve and glCopyImageSubData
+// between them crosses a size-incompatible pair, which vkCmdCopyImage leaves undefined
+// (VUID-vkCmdCopyImage-srcImage-01548). The renderbuffer path used to carry a hand-maintained second
+// copy of this table that was missing four rows - RGBA2, RGBA4, RGB5A1 and RGBA12 - so those four
+// renderbuffer formats either got no image at all or a 16-bit-packed one facing a 32-bit texture.
+struct TextureFormatInfo {
+    VkFormat format = VK_FORMAT_UNDEFINED;
+    // The GL format has three channels and is carried in a four-channel image; a shadow upload has
+    // to be expanded, inserting `alphaBytes` after every `componentByteCount * 3` source bytes.
+    Bool expandRgbToRgba = false;
+    Uint32 componentByteCount = 0;
+    Array<Uint8, 4> alphaBytes = {0, 0, 0, 0};
+};
+
+// Callers that only need the backing VkFormat (a renderbuffer has no shadow upload to reshape) take
+// `.format` and ignore the rest.
+TextureFormatInfo ResolveTextureFormatInfo(TextureInternalFormat format);
+
+// A GL 1D-ARRAY level keeps its LAYER COUNT in the state-side HEIGHT: that is what
+// glTexImage2D(GL_TEXTURE_1D_ARRAY, width, layers) means, and the frontend records the level
+// as {width, layers, 1} (see GL_Texture.cpp's AllocateStorage and the completeness walk in
+// TextureObject.cpp, which shrinks only x down the chain). Vulkan packs it the other way: a
+// 1D array is a VK_IMAGE_TYPE_1D image whose extent.height MUST be 1 and whose layers live in
+// arrayLayers - i.e. in the slot this backend reads out of z. So every place that turns a GL
+// level size into Vulkan image geometry has to move the count across first, and every GL-space
+// sub-box that rides along with it has to move its y the same way. DirectGLES performs the
+// identical remap onto the ES 2D array it maps 1D arrays to (GetBackendUploadSize).
+//
+// Applied to nothing else: a 2D array, a cube array and a 3D texture all already carry their
+// depth/layer count in z, which is where the Vulkan side expects it.
+inline IntVec3 ToVulkanLevelExtent(TextureTarget stateTarget, const IntVec3& glTexelSize) {
+    if (stateTarget == TextureTarget::Texture1DArray) {
+        return {glTexelSize.x(), 1, glTexelSize.y()};
+    }
+    return glTexelSize;
+}
+
+// How many Vulkan array layers (or, for a 3D image, z slices) a GL framebuffer attachment spans.
+//
+// THE ONE COPY, deliberately. This used to exist twice - privately in VkRenderPassManager.cpp and
+// again in VkClearManager.cpp - and the two are not independent: the render pass builds the
+// attachment view and VkFramebufferCreateInfo::layers from one, while the CLEAR key built from the
+// other is written verbatim into VkImageSubresourceRange::layerCount when a queued glClear is
+// materialised outside a render pass (MaterializePendingClearForTexture). They are two consumers
+// of the same GL clear, so any disagreement means the same glClear produces two different pictures
+// depending only on which path happens to consume it first - and the materialise path then POPS
+// the entry, so the other one never runs. Fixing one copy and leaving the other is exactly how
+// that split gets introduced; keep them the same function.
+//
+// Two shapes make this more than `size.z()`:
+//   * GL_TEXTURE_1D_ARRAY keeps its layer count in the state-side HEIGHT (see ToVulkanLevelExtent
+//     just above), so z reads 1 and every layer above the first was silently dropped.
+//   * GL_TEXTURE_CUBE_MAP is attached layered as its REPRESENTATIVE upload target, the +X face
+//     (ResolveRepresentableFramebufferTextureUploadTarget), and one face's level size has z = 1 -
+//     but a layered cube attachment names all six faces (GL 4.6 core 9.2.8), which are the image's
+//     six array layers. A cube ARRAY needs no such arm: its representative target carries 6n in z.
+inline Uint32 ResolveAttachmentLayerCount(const MG_State::GLState::FramebufferAttachmentObject& attachment) {
+    if (!attachment.IsLayered()) {
+        return 1u;
+    }
+    const auto& texture = attachment.GetTexture();
+    const TextureTarget target = texture != nullptr ? texture->GetTarget() : TextureTarget::Unknown;
+    if (target == TextureTarget::TextureCubeMap) {
+        return 6u;
+    }
+    return static_cast<Uint32>(std::max(ToVulkanLevelExtent(target, attachment.GetSize()).z(), 1));
+}
+
+// A GL framebuffer attachment's level/layer, and a GL image unit's, are relative to the texture
+// the application NAMED. When that texture was created by glTextureView (ARB_texture_view) they
+// are relative to the VIEW, and have to be shifted into the storage image's numbering before they
+// can index a Vulkan subresource - DirectVulkan gives a view no image of its own, it shares the
+// storage texture's (VkTextureManager::StorageTextureOf).
+//
+// Apply EXACTLY ONCE, at the boundary where a GL level/layer becomes a subresource index. Every
+// GetOrCreate*View entry point below expects values that have already been through here, and so
+// does everything that reads or copies an attachment directly. Both are identity on a plain
+// texture (TEXTURE_VIEW_MIN_LEVEL / MIN_LAYER are 0 there), so the conversion is unconditional
+// and there is no second, view-only code path to keep in step.
+inline Uint32 ToStorageMipLevel(const MG_State::GLState::ITextureObject* texture, Int glLevel) {
+    const Uint32 level = static_cast<Uint32>(glLevel > 0 ? glLevel : 0);
+    return texture != nullptr ? level + static_cast<Uint32>(texture->GetViewMinLevel()) : level;
+}
+
+inline Uint32 ToStorageArrayLayer(const MG_State::GLState::ITextureObject* texture, Int glLayer) {
+    const Uint32 layer = static_cast<Uint32>(glLayer > 0 ? glLayer : 0);
+    return texture != nullptr ? layer + static_cast<Uint32>(texture->GetViewMinLayer()) : layer;
+}
+
 class VkTextureManager {
 public:
     // Monotonic epoch bumped whenever a texture VkImage is (re)created. The render-pass
     // manager keys its per-draw fast path on this so an attachment's image recreation
     // invalidates the cached render pass (dirty-flag tracking; portable to Vulkan 1.1).
     Uint64 GetTextureImageEpoch() const { return m_textureImageEpoch; }
+    // Bumped whenever any tracked texture resource is erased; cached
+    // TextureResource pointers are valid only while this is unchanged.
+    Uint64 GetResourceEraseEpoch() const { return m_resourceEraseEpoch; }
 
     struct TextureIdentity {
         MG_State::GLState::ITextureObject* texture = nullptr;
@@ -52,6 +154,20 @@ public:
         VkCommandPool commandPool = VK_NULL_HANDLE;
         VkQueue graphicsQueue = VK_NULL_HANDLE;
         Uint32 frameCount = 0;
+        // 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 {
@@ -60,12 +176,16 @@ public:
             Uint32 baseArrayLayer = 0;
             Uint32 layerCount = 1;
             VkImageViewType viewType = VK_IMAGE_VIEW_TYPE_2D;
+            // May differ from the image format: sRGB images attach through their UNORM
+            // twin while GL_FRAMEBUFFER_SRGB is disabled.
+            VkFormat viewFormat = VK_FORMAT_UNDEFINED;
 
             Bool operator==(const AttachmentViewKey& other) const {
                 return mipLevel == other.mipLevel &&
                        baseArrayLayer == other.baseArrayLayer &&
                        layerCount == other.layerCount &&
-                       viewType == other.viewType;
+                       viewType == other.viewType &&
+                       viewFormat == other.viewFormat;
             }
         };
 
@@ -76,6 +196,8 @@ public:
                 hash ^= std::hash<Uint32>{}(key.layerCount) + 0x9e3779b9u + (hash << 6) + (hash >> 2);
                 hash ^= std::hash<Uint32>{}(static_cast<Uint32>(key.viewType)) +
                         0x9e3779b9u + (hash << 6) + (hash >> 2);
+                hash ^= std::hash<Uint32>{}(static_cast<Uint32>(key.viewFormat)) +
+                        0x9e3779b9u + (hash << 6) + (hash >> 2);
                 return hash;
             }
         };
@@ -96,17 +218,35 @@ public:
             }
         };
 
+        // Layer range and aspect join the key because a GL texture view (ARB_texture_view) can
+        // differ from its storage on either: the Better Clouds shape samples ONE D24S8 image
+        // through two GL names in one draw, the parent with the stencil aspect and the view with
+        // the depth aspect, and a layer-sliced view of an array texture names a sub-range of the
+        // same image. Without these two fields those views would alias each other in the cache.
         struct SampledImageViewKey {
             Uint32 baseMipLevel = 0;
             Uint32 levelCount = 1;
+            Uint32 baseArrayLayer = 0;
+            Uint32 layerCount = 1;
             VkImageViewType viewType = VK_IMAGE_VIEW_TYPE_2D;
             VkFormat format = VK_FORMAT_UNDEFINED;
+            VkImageAspectFlags aspect = VK_IMAGE_ASPECT_COLOR_BIT;
+            // GL_TEXTURE_SWIZZLE_* is per-texture state, so two views over one storage with the
+            // same window but different swizzles are different views. Baked into the key because
+            // a GL texture view's ONLY sampled view lives in this cache: unlike the storage
+            // texture's own sampledView, which SyncTextureViews rebuilds whenever the params
+            // version moves, nothing else would ever notice a swizzle change on a view.
+            Uint32 componentSwizzle = 0;
 
             Bool operator==(const SampledImageViewKey& other) const {
                 return baseMipLevel == other.baseMipLevel &&
                        levelCount == other.levelCount &&
+                       baseArrayLayer == other.baseArrayLayer &&
+                       layerCount == other.layerCount &&
                        viewType == other.viewType &&
-                       format == other.format;
+                       format == other.format &&
+                       aspect == other.aspect &&
+                       componentSwizzle == other.componentSwizzle;
             }
         };
 
@@ -114,10 +254,15 @@ public:
             SizeT operator()(const SampledImageViewKey& key) const {
                 SizeT hash = std::hash<Uint32>{}(key.baseMipLevel);
                 hash ^= std::hash<Uint32>{}(key.levelCount) + 0x9e3779b9u + (hash << 6) + (hash >> 2);
+                hash ^= std::hash<Uint32>{}(key.baseArrayLayer) + 0x9e3779b9u + (hash << 6) + (hash >> 2);
+                hash ^= std::hash<Uint32>{}(key.layerCount) + 0x9e3779b9u + (hash << 6) + (hash >> 2);
                 hash ^= std::hash<Uint32>{}(static_cast<Uint32>(key.viewType)) +
                         0x9e3779b9u + (hash << 6) + (hash >> 2);
                 hash ^= std::hash<Uint32>{}(static_cast<Uint32>(key.format)) +
                         0x9e3779b9u + (hash << 6) + (hash >> 2);
+                hash ^= std::hash<Uint32>{}(static_cast<Uint32>(key.aspect)) +
+                        0x9e3779b9u + (hash << 6) + (hash >> 2);
+                hash ^= std::hash<Uint32>{}(key.componentSwizzle) + 0x9e3779b9u + (hash << 6) + (hash >> 2);
                 return hash;
             }
         };
@@ -156,7 +301,25 @@ public:
         VkImageViewType viewType = VK_IMAGE_VIEW_TYPE_2D;
         VkSampleCountFlagBits sampleCount = VK_SAMPLE_COUNT_1_BIT;
         VkImageCreateFlags imageCreateFlags = 0;
+        // Usage the live image was created with. STORAGE is only requested for textures that
+        // have actually been bound to a GL image unit, because on Adreno a storage-capable
+        // image loses UBWC bandwidth compression; a later image binding upgrades the usage
+        // and recreates the image, so the resolved usage has to be part of the compatibility
+        // check that decides whether the existing image can be kept.
+        VkImageUsageFlags usageFlags = 0;
+        // True once this image was (re)resolved while the texture was already marked as an
+        // image-unit texture. Distinguishes "not upgraded yet" from "cannot be upgraded"
+        // (a format whose optimalTilingFeatures lack STORAGE_IMAGE never gains the bit), so
+        // NeedsStorageImagePreparation cannot ask for a recreate that will never happen.
+        Bool storageUsageResolved = false;
         Uint16 syncedTextureParamsVersion = 0;
+        // Recording generation (VkTextureManager::GetRecordingGeneration) of the last
+        // command referencing this image that was recorded into the CURRENT frame
+        // command buffer. An image untouched by the open recording may have its
+        // out-of-pass work (deferred clears, sampled-layout transitions) recorded
+        // into the frame's PRE command buffer - which executes strictly before the
+        // frame's commands - instead of splitting the active render pass.
+        Uint64 lastRecordingGeneration = 0;
         // Snapshot of ITextureObject::GetContentVersion() at the last successful sync;
         // lets SyncTexture skip the whole re-check/re-upload when content is unchanged.
         Uint64 syncedContentVersion = 0;
@@ -164,6 +327,12 @@ public:
         // as defense-in-depth: any path that grows the level set (which resizes the sampled view)
         // busts the skip even if it failed to bump the content version.
         Uint32 syncedMipLevelCount = 0;
+        // Snapshot of ITextureObject::GetShapeVersion() at the last successful sync. The content
+        // version alone does NOT cover a re-specification: glTexImage2D(..., nullptr) on an
+        // already-defined level changes its size or format and dirties no texel, so it moves the
+        // shape version and nothing else. Without this in the early-out key the image, its views
+        // and therefore imageSize() all keep answering with the texture's PREVIOUS shape.
+        Uint64 syncedShapeVersion = 0;
 
         TextureResource() = default;
         TextureResource(const TextureResource&) = delete;
@@ -189,9 +358,13 @@ public:
             std::swap(this->viewType, that.viewType);
             std::swap(this->sampleCount, that.sampleCount);
             std::swap(this->imageCreateFlags, that.imageCreateFlags);
+            std::swap(this->usageFlags, that.usageFlags);
+            std::swap(this->storageUsageResolved, that.storageUsageResolved);
             std::swap(this->syncedTextureParamsVersion, that.syncedTextureParamsVersion);
+            std::swap(this->lastRecordingGeneration, that.lastRecordingGeneration);
             std::swap(this->syncedContentVersion, that.syncedContentVersion);
             std::swap(this->syncedMipLevelCount, that.syncedMipLevelCount);
+            std::swap(this->syncedShapeVersion, that.syncedShapeVersion);
         }
 
         void Reset() {
@@ -250,9 +423,12 @@ public:
             viewType = VK_IMAGE_VIEW_TYPE_2D;
             sampleCount = VK_SAMPLE_COUNT_1_BIT;
             imageCreateFlags = 0;
+            usageFlags = 0;
+            storageUsageResolved = false;
             syncedTextureParamsVersion = 0;
             syncedContentVersion = 0;
             syncedMipLevelCount = 0;
+            syncedShapeVersion = 0;
         }
 
         ~TextureResource() {
@@ -263,9 +439,78 @@ 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.
+    void CollectAllDeferredReleases();
+
+    // ---- GL texture views (ARB_texture_view / GL 4.6 core 8.18) ----
+    // The GL texture whose STORAGE backs the given one: itself, or - for a texture created by
+    // glTextureView - the texture it views. Every image-scoped question (which VkImage, its
+    // LAYOUT, its uploads, its extent, its usage) must be asked of this object, because a view
+    // has none of its own; only the VkImageViews differ per GL texture object. Sharing one
+    // TextureResource is not an optimisation, it is the only correct arrangement: layout is a
+    // property of the image, and VulkanRenderer caches raw pointers straight to the resource's
+    // layout field, so a second resource aliasing the same image would desynchronise the moment
+    // either of them transitioned it.
+    static MG_State::GLState::ITextureObject& StorageTextureOf(MG_State::GLState::ITextureObject& texture);
+
+    // The window a GL texture object opens onto its storage image. For a plain texture this is
+    // the resource's own full extent; for a view it is the sub-range, format and aspect
+    // glTextureView gave it. Views built from a non-default window must live in the KEYED caches
+    // (attachmentViews / alternateSampledViews), never in the per-mip vectors, which belong to
+    // the storage texture's own defaults.
+    struct TextureViewWindow {
+        Uint32 baseMipLevel = 0;
+        Uint32 levelCount = 1;
+        Uint32 baseArrayLayer = 0;
+        Uint32 layerCount = 1;
+        VkFormat format = VK_FORMAT_UNDEFINED;
+        VkImageViewType viewType = VK_IMAGE_VIEW_TYPE_2D;
+        VkImageAspectFlags sampledAspect = VK_IMAGE_ASPECT_COLOR_BIT;
+        VkComponentMapping components{VK_COMPONENT_SWIZZLE_R, VK_COMPONENT_SWIZZLE_G, VK_COMPONENT_SWIZZLE_B,
+                                      VK_COMPONENT_SWIZZLE_A};
+        Bool isTextureView = false;
+    };
+
+    // The four component swizzles packed into one value, for the sampled-view cache key.
+    static Uint32 PackComponentSwizzle(const VkComponentMapping& components) {
+        return (static_cast<Uint32>(components.r) & 0xFFu) | ((static_cast<Uint32>(components.g) & 0xFFu) << 8) |
+               ((static_cast<Uint32>(components.b) & 0xFFu) << 16) |
+               ((static_cast<Uint32>(components.a) & 0xFFu) << 24);
+    }
+    TextureViewWindow ResolveTextureViewWindow(MG_State::GLState::ITextureObject& texture,
+                                               const TextureResource& resource) const;
+    // Records what a GL texture view needs of the image it views, so the next sync of the
+    // STORAGE texture creates (or recreates and copies forward) an image the view can be built
+    // over. See m_viewRequestedImageFlags for why this is lazy rather than unconditional.
+    void NoteTextureViewImageRequirements(MG_State::GLState::ITextureObject& viewTexture,
+                                          MG_State::GLState::ITextureObject& storageTexture);
+    VkImageCreateFlags GetViewRequestedImageFlags(const MG_State::GLState::ITextureObject& storageTexture) const;
+    // Appends every format a GL texture view reinterprets this storage as, for the narrowed
+    // VkImageFormatListCreateInfo the image is created with.
+    void AppendViewRequestedFormats(const MG_State::GLState::ITextureObject& storageTexture,
+                                    Vector<VkFormat>& outFormats) const;
+    // Builds (and caches, keyed by the whole window) one sampled VkImageView over a storage
+    // image. Shared back end of every GL-texture-view sampled path.
+    VkImageView GetOrCreateWindowedSampledView(MG_State::GLState::ITextureObject& texture,
+                                               TextureResource& resource, const TextureViewWindow& window);
 
     TextureResource* SyncTextureAndGetDescriptor(
         MG_State::GLState::ITextureObject& texture);
@@ -284,18 +529,76 @@ 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);
 
-    static VkImageAspectFlags ResolveSampledImageViewAspectMask(VkImageAspectFlags imageAspect);
+    // Recording-generation bookkeeping for the pre-pass command stream. The
+    // generation advances every time the frame command buffer (re)begins
+    // recording; a resource whose stamp does not match was not referenced by
+    // any command in the open recording, so its out-of-pass work may safely
+    // execute ahead of the whole recording (in the pre command buffer).
+    void AdvanceRecordingGeneration() { ++m_recordingGeneration; }
+    void StampResourceRecordingUse(TextureResource& resource) const {
+        resource.lastRecordingGeneration = m_recordingGeneration;
+    }
+    // Map-lookup variant for callers that only hold the GL texture object.
+    void StampTextureRecordingUse(MG_State::GLState::ITextureObject* texture);
+    Bool WasTouchedThisRecording(const TextureResource& resource) const {
+        return resource.lastRecordingGeneration == m_recordingGeneration;
+    }
+    // Records that this texture is bound to a GL image unit, so its image must carry
+    // VK_IMAGE_USAGE_STORAGE_BIT. Must be called before NeedsStorageImagePreparation, and
+    // therefore before the render pass is committed: an image that has to be upgraded is
+    // recreated, which is illegal inside a render pass. Sticky for the texture's lifetime -
+    // GL lets an image binding come and go, and re-creating the image every time it does
+    // would cost far more than the compression it wins back.
+    void MarkStorageImageTexture(MG_State::GLState::ITextureObject& texture);
+    // True when this texture is marked but its live image predates the mark, i.e. the next sync
+    // will recreate it with STORAGE usage and copy the old contents forward. Callers use this to
+    // submit their pending recording first, so that copy cannot read pre-flush content.
+    Bool NeedsStorageUsageUpgrade(MG_State::GLState::ITextureObject& texture) const;
+    // The same ordering question for the other recreate-and-preserve trigger: true when this
+    // texture's live image carries a shorter mip chain than a full one, so defining the missing
+    // levels recreates it and copies the old contents forward.
+    Bool NeedsMipChainGrowth(MG_State::GLState::ITextureObject& texture) const;
+    // Non-mutating probe for the per-draw storage-image fast path: true when preparing this
+    // texture as a storage image may need work that is illegal inside a render pass (resource
+    // creation, dirty-content upload, or a layout transition to GENERAL). Unknown state reports
+    // true - a false positive merely ends the render pass, a false negative would skip a barrier.
+    Bool NeedsStorageImagePreparation(MG_State::GLState::ITextureObject& texture) const;
+
+    // `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 layerCount = 1);
+                               Uint32 baseMipLevel = 0, Uint32 levelCount = 1);
 
     SizeT CollectGarbage();
 
@@ -325,6 +628,9 @@ public:
 private:
     // Bumped in SyncTextureResource right after vmaCreateImage(texture). See GetTextureImageEpoch().
     Uint64 m_textureImageEpoch = 1;
+    // See AdvanceRecordingGeneration. Starts above every resource's default
+    // stamp of 0 so a fresh resource counts as untouched.
+    Uint64 m_recordingGeneration = 1;
 
     Bool SyncTexture(MG_State::GLState::ITextureObject &texture,
                      TextureResource &outResource);
@@ -355,18 +661,31 @@ private:
     void DeferViewRelease(VkImageView view);
     void CollectDeferredReleases(Uint32 frameIndex);
     void DestroyDeferredReleases();
+    // Frees the fence/command buffer/staging buffer of every in-flight texture
+    // upload whose fence has signaled (submission order = completion order on
+    // the single queue, so the scan stops at the first still-pending entry).
+    // waitAll blocks on every entry - Shutdown's drain.
+    void ReclaimCompletedUploads(Bool waitAll = false);
     static TextureIdentity MakeTextureIdentity(MG_State::GLState::ITextureObject* texture);
     void EraseTrackedTexture(const TextureIdentity& identity);
     void PruneStaleTextureAliases(MG_State::GLState::ITextureObject* texture);
+    SizeT PruneDeadTextures();
 
     VkDevice m_device = VK_NULL_HANDLE;
     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;
 
     Uint8 m_gcCounter = 0;
+    // Frame-boundary GC gate: counts BeginFrame calls, not draws, so texture churn
+    // through non-draw paths (FBO clears, readbacks) still reaches the prune.
+    Uint32 m_gcFrameCounter = 0;
     // Active only between BeginDrawSyncScope/EndDrawSyncScope; identities of
     // textures already fully synced in the current draw (small N -> flat scan).
     Bool m_drawSyncScopeActive = false;
@@ -379,9 +698,105 @@ private:
         TextureResource* resource = nullptr;
     };
     Vector<DrawSyncedTexture> m_drawSyncedThisDraw;
+    // Cross-draw sampled-texture memo: the same few textures (atlas, lightmap)
+    // are resolved on every draw, so cache their resource pointers and skip the
+    // alive/resource map lookups. Node-based std::unordered_map keeps the
+    // pointees stable across inserts; erases bump m_resourceEraseEpoch, which
+    // every memo entry must match. SyncTexture still runs on memo hits, so
+    // content/param freshness is unaffected. A dead-then-reused texture address
+    // cannot false-hit: the new object carries a new lifetime id.
+    struct SyncedTextureMemoEntry {
+        const MG_State::GLState::ITextureObject* texture = nullptr;
+        Uint64 lifetimeId = 0;
+        Uint64 eraseEpoch = 0;
+        TextureResource* resource = nullptr;
+    };
+    static constexpr Uint32 kSyncedTextureMemoSize = 8;
+    SyncedTextureMemoEntry m_syncedTextureMemo[kSyncedTextureMemoSize];
+    Uint32 m_syncedTextureMemoNext = 0;
+    Uint64 m_resourceEraseEpoch = 1;
+    // 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).
+    std::unordered_set<TextureIdentity, TextureIdentityHash> m_storageImageTextures;
+    // Extra VkImageCreateFlags a GL texture view needs on the storage image it views, keyed by
+    // the STORAGE texture's identity. Requested lazily, exactly like STORAGE usage above and for
+    // the same reason: VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT costs bandwidth compression on tilers
+    // (it is what VK_KHR_image_format_list exists to claw back), so setting it on every
+    // immutable-storage texture would tax every glTexStorage2D render target in a game for a
+    // feature almost none of them use. A SAME-format view - which is the common case, and the
+    // Better Clouds case - needs no flag at all and therefore costs nothing.
+    std::unordered_map<TextureIdentity, VkImageCreateFlags, TextureIdentityHash> m_viewRequestedImageFlags;
+    // Every VkFormat a GL texture view has asked to reinterpret this storage as. The narrowed
+    // VkImageFormatListCreateInfo the image is created with must name them: the list is a promise
+    // that NO other format will ever be viewed, and building a view outside it is
+    // VUID-VkImageViewCreateInfo-pNext-01585. Keyed, like the flags above, by the STORAGE texture.
+    std::unordered_map<TextureIdentity, std::unordered_set<VkFormat>, TextureIdentityHash> m_viewRequestedFormats;
+    // Supported multisample counts per format, so repeat texture syncs do not
+    // re-query vkGetPhysicalDeviceImageFormatProperties.
+    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). Each flushed batch's
+    // 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;
+        Vector<UploadStagingBlock> stagingBlocks;
+    };
+    Vector<PendingUploadReclaim> m_pendingUploadReclaims;
 };
 } // namespace MobileGL::MG_Backend::DirectVulkan

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/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.MagmaDisableSubgroup;
+    }
+
+    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.MagmaFixIterationRPSubgroupScratch !=
+               MG_Config::QuirkOverride::ForceOff;
+    }
+
+    inline Bool ShouldFixIterationRPBarrier() {
+        return MG_Config::Features.MagmaIterationRPFixBarrier;
+    }
+
+    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.MagmaDeriveNumSubgroups != MG_Config::QuirkOverride::ForceOff;
+    }
+} // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/VkIncludes.h

@@ -52,19 +52,60 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     }
 } // namespace MobileGL::MG_Backend::DirectVulkan
 
+namespace MobileGL::MG_Backend::DirectVulkan {
+    // GL renders into sRGB color attachments RAW while GL_FRAMEBUFFER_SRGB is disabled
+    // (the core-profile default); Vulkan sRGB attachments always encode on write. The
+    // attachment view (and render pass format) therefore drops to the UNORM twin
+    // whenever the capability is off. Sampled views keep the sRGB format (decode on
+    // sample is unconditional in GL).
+    inline VkFormat ResolveSrgbAttachmentWriteFormat(VkFormat format, bool framebufferSrgbEnabled) {
+        if (framebufferSrgbEnabled) return format;
+        switch (format) {
+        case VK_FORMAT_R8G8B8A8_SRGB:
+            return VK_FORMAT_R8G8B8A8_UNORM;
+        case VK_FORMAT_B8G8R8A8_SRGB:
+            return VK_FORMAT_B8G8R8A8_UNORM;
+        default:
+            return format;
+        }
+    }
+} // 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);                                                                           \
         if (_vk_verify_result != VK_SUCCESS) {                                                                         \
-            MGLOG_F("Vulkan error %s (%d) at %s:%d" __VA_OPT__(" - ") __VA_ARGS__,                                  \
+            __VA_OPT__(MGLOG_F(__VA_ARGS__);)                                                                          \
+            MGLOG_F("Vulkan error %s (%d) at %s:%d",                                                                   \
                     MobileGL::MG_Backend::DirectVulkan::VkResultToString(_vk_verify_result),                          \
                     _vk_verify_result, __FILE__, __LINE__);                                                            \
         }                                                                                                              \
-        MOBILEGL_ASSERT(_vk_verify_result == VK_SUCCESS, "Vulkan error %s (%d) at %s:%d" __VA_OPT__(" - ") __VA_ARGS__, MobileGL::MG_Backend::DirectVulkan::VkResultToString(_vk_verify_result), _vk_verify_result, __FILE__, __LINE__);  \
+        MOBILEGL_ASSERT(_vk_verify_result == VK_SUCCESS, "Vulkan error %s (%d) at %s:%d",                              \
+                        MobileGL::MG_Backend::DirectVulkan::VkResultToString(_vk_verify_result),                       \
+                        _vk_verify_result, __FILE__, __LINE__);                                                        \
     } while (0)
 
 #define XXHASH_VERIFY(expr, ...)                                                                                       \
     do {                                                                                                               \
         XXH_errorcode _xxh_verify_result = (expr);                                                                     \
-        MOBILEGL_ASSERT(_xxh_verify_result == XXH_OK, "XXHash error %d at %s:%d" __VA_OPT__(" - ") __VA_ARGS__, _xxh_verify_result, __FILE__, __LINE__);  \
+        if (_xxh_verify_result != XXH_OK) {                                                                            \
+            __VA_OPT__(MGLOG_F(__VA_ARGS__);)                                                                          \
+        }                                                                                                              \
+        MOBILEGL_ASSERT(_xxh_verify_result == XXH_OK, "XXHash error %d at %s:%d", _xxh_verify_result, __FILE__,        \
+                        __LINE__);                                                                                     \
     } while (0)

MobileGL/MG_Benchmark/CMakeLists.txt

@@ -41,4 +41,8 @@ add_test(NAME SanityBench COMMAND SanityBench --benchmark_counters_tabular=true)
 set_tests_properties(SanityBench PROPERTIES LABELS benchmark)
 
 add_subdirectory(Program)
-add_subdirectory(Buffer)
+add_subdirectory(Buffer)
+add_subdirectory(Driver)
+add_subdirectory(Container)
+add_subdirectory(ShaderCache)
+add_subdirectory(Transpile)

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_Benchmark/ShaderCache/CMakeLists.txt

@@ -0,0 +1,21 @@
+cmake_minimum_required(VERSION 3.24)
+
+add_executable(
+        TranslationCacheBench
+        TranslationCacheBench.cpp
+)
+
+target_include_directories(TranslationCacheBench PRIVATE
+        ${MGL_ROOT}/include
+        ${MGL_ROOT}/MobileGL
+        ${MGL_ROOT}/3rdparty/SPIRV-Reflect
+)
+
+target_link_libraries(
+        TranslationCacheBench PRIVATE
+        benchmark::benchmark
+        ${LINK_LIBRARIES}
+)
+
+add_test(NAME TranslationCacheBench COMMAND TranslationCacheBench --benchmark_counters_tabular=true)
+set_tests_properties(TranslationCacheBench PROPERTIES LABELS benchmark)

MobileGL/MG_Benchmark/ShaderCache/TranslationCacheBench.cpp

@@ -0,0 +1,457 @@
+// MobileGL - MobileGL/MG_Benchmark/ShaderCache/TranslationCacheBench.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
+
+// What the two-level shader translation memo is worth, measured on the workload that
+// motivated it: the KHR-GL33.texture_swizzle.smoke_* shape, where one case builds 2592
+// programs out of a handful of distinct sources.
+//
+// Four pairs of cases, each Off/On:
+//
+//   ProgramLink      - the whole glCompileShader + glLinkProgram path for one program, with
+//                      FRESH SHADER OBJECTS every iteration. This is the CTS shape exactly,
+//                      and it is the headline case now. It used to be the PESSIMISTIC one:
+//                      a hit still paid for both glslang parses, because the parse happens
+//                      at glCompileShader - a different entry point from the one L1
+//                      memoizes - and fresh shader objects meant ShaderCompileAdoptionMap
+//                      could not hand the earlier parse over either. L1c is what closed
+//                      that: the compile half of the memo recognises each stage's source
+//                      and publishes its verdict without parsing, so on a hit this case now
+//                      constructs no glslang object at all.
+//
+//   SharedShaderLink - the same program population with the shader objects KEPT ALIVE, so
+//                      the parses happen once outside the measured loop whatever the cache
+//                      does. That makes it the CONTROL for L1c rather than a target: its
+//                      numbers should not move, and if they do, L1c has added cost to a
+//                      path it was supposed to leave alone.
+//
+//   DeferredParseLink - the shape where L1c could LOSE: a constant vertex source (which
+//                      hits L1c and therefore skips its parse) against a fresh fragment
+//                      source every iteration (which makes the PROGRAM key miss, so the
+//                      skipped parse has to happen inside the link after all). Same parse
+//                      count either way, so the pair should land within noise; see its own
+//                      header below.
+//
+//   EsslTranspile    - the DirectGLES backend segment: the SPIR-V pass chain plus
+//                      SPIRV-Cross. Runs the driver-INDEPENDENT half of the real chain (the
+//                      passes SyncToBackend runs unconditionally, plus the two stage-gated
+//                      ones a fragment module reaches) so the miss path costs what
+//                      production costs; the capability-gated passes need a live ES driver
+//                      and are not reachable from a benchmark process.
+//
+// Every On case runs with a warm cache: the first iteration misses and every one after it
+// hits, which is exactly the steady state of a 2592-program smoke case.
+
+#include <benchmark/benchmark.h>
+
+#include <string>
+
+#include "Config.h"
+#include "Includes.h"
+#include "Init.h"
+#include "MG_Impl/GLImpl/Program/GL_Program.h"
+#include "MG_State/GLState/Core.h"
+#include "MG_State/GLState/ProgramState/ProgramTranslationCache.h"
+#include "MG_Util/ShaderTranspiler/ShaderCompiler.h"
+#include "MG_Util/ShaderTranspiler/SpvcSession.h"
+#include "MG_Util/ShaderTranspiler/TranslationCache.h"
+#include "MG_Util/ShaderTranspiler/Types.h"
+
+using namespace MobileGL;
+using namespace MobileGL::MG_Util::ShaderTranspiler;
+
+namespace {
+    const char* kVertexSource = R"(#version 460
+layout(location = 0) in vec3 aPos;
+out vec3 vPos;
+out vec2 vUv;
+void main() {
+    vPos = aPos;
+    vUv = aPos.xy * 0.5 + 0.5;
+    gl_Position = vec4(aPos, 1.0);
+}
+)";
+
+    // Shaped after gl3cTextureSwizzleTests.cpp's template: a sampler of one type, one
+    // TEXTURE_ACCESS, one CHANNEL, and an output whose BASIC_TYPE is the only thing that
+    // varies within a case. Padded with enough real arithmetic that the translation chain
+    // is doing work rather than measuring fixed overheads.
+    // `padLines` = 0 is the honest CTS size: gl3cTextureSwizzleTests' smoke template is a
+    // handful of lines, and that is the workload the memo exists for. The padded variant is
+    // kept alongside it because a shaderpack stage is orders of magnitude bigger, and the
+    // two bracket the ratio the cache is worth in practice.
+    String SwizzleLikeFragment(const String& prefix, const int padLines) {
+        String source = "#version 460\n";
+        source += "in vec3 vPos;\n";
+        source += "in vec2 vUv;\n";
+        source += "layout(location = 0) out " + prefix + "vec4 fragColor;\n";
+        source += "uniform sampler2D uTex;\n";
+        source += "uniform vec4 uTint;\n";
+        source += "uniform mat4 uModel;\n";
+        source += "uniform float uArr[8];\n";
+        source += "void main() {\n";
+        source += "    vec4 s = texture(uTex, vUv);\n";
+        source += "    float acc = s.r;\n";
+        for (int i = 0; i < padLines; ++i) {
+            source += "    acc = acc * 1.0001 + sin(acc + " + std::to_string(i) + ".0) * cos(acc);\n";
+        }
+        source += "    for (int i = 0; i < 8; ++i) acc += uArr[i];\n";
+        source += "    vec4 p = uModel * vec4(vPos, 1.0);\n";
+        source += "    fragColor = " + prefix + "vec4((s + uTint) * acc + p);\n";
+        source += "}\n";
+        return source;
+    }
+
+    class CacheModeScope {
+    public:
+        explicit CacheModeScope(const Bool enabled)
+            : m_saved(MG_Config::Features.ShaderTranslationCache) {
+            MG_Config::Features.ShaderTranslationCache =
+                enabled ? MG_Config::QuirkOverride::ForceOn : MG_Config::QuirkOverride::ForceOff;
+        }
+        ~CacheModeScope() { MG_Config::Features.ShaderTranslationCache = m_saved; }
+
+    private:
+        const MG_Config::QuirkOverride m_saved;
+    };
+
+    class SyncCompileScope {
+    public:
+        SyncCompileScope() : m_saved(MG_Config::Features.AsyncShaderCompile) {
+            MG_Config::Features.AsyncShaderCompile = MG_Config::QuirkOverride::ForceOff;
+        }
+        ~SyncCompileScope() { MG_Config::Features.AsyncShaderCompile = m_saved; }
+
+    private:
+        const MG_Config::QuirkOverride m_saved;
+    };
+
+    // One program, built the way the CTS builds one: fresh shader objects every time.
+    void LinkOneProgram(const String& vertexSource, const String& fragmentSource) {
+        using namespace MG_Impl::GLImpl;
+        const GLuint vs = CreateShader(GL_VERTEX_SHADER);
+        const char* vsText = vertexSource.c_str();
+        ShaderSource(vs, 1, &vsText, nullptr);
+        CompileShader(vs);
+
+        const GLuint fs = CreateShader(GL_FRAGMENT_SHADER);
+        const char* fsText = fragmentSource.c_str();
+        ShaderSource(fs, 1, &fsText, nullptr);
+        CompileShader(fs);
+
+        const GLuint program = CreateProgram();
+        AttachShader(program, vs);
+        AttachShader(program, fs);
+        LinkProgram(program);
+        benchmark::DoNotOptimize(program);
+
+        DeleteProgram(program);
+        DeleteShader(vs);
+        DeleteShader(fs);
+    }
+
+    Vector<Uint32> BuildSanitizedFragmentSpirv(const String& fragmentSource) {
+        ShaderAttrib attrib{.shaderType = GL_FRAGMENT_SHADER, .sourceStr = fragmentSource};
+        auto shader = ShaderCompiler::CompileShader(attrib);
+        if (!shader) return {};
+        ProgramAttrib programAttrib{.shaders = {shader.value()}};
+        auto program = ShaderCompiler::LinkProgram(programAttrib);
+        if (!program) return {};
+        ProgramBinaryAttrib binaryAttrib{.shaderTypes = {GL_FRAGMENT_SHADER}, .program = *program.value()};
+        auto binary = ShaderCompiler::GetSpirvBinaryFromProgram(binaryAttrib);
+        if (!binary || binary->empty()) return {};
+        Vector<Uint32> sanitized;
+        if (!ShaderCompiler::SanitizeAndOptimizeBinary(binary->front(), sanitized)) return {};
+        return sanitized;
+    }
+
+    // The driver-independent part of BackendProgramObjectImpl::TranspileSpirvToEssl, in the
+    // same order. What is missing is only the capability-gated passes (viewport lowering,
+    // multisample clamping, noperspective emulation, the image-format bake), which cannot
+    // fire without a live ES driver to arm them.
+    Bool TranspileLikeDirectGles(const Vector<Uint32>& spirv, const Uint esslVersion, String& outEssl) {
+        Vector<Uint32> a;
+        const Vector<Uint32>* effective = &spirv;
+        if (ShaderCompiler::StripUboMemberRelaxedPrecisionForEssl(*effective, a, false) && !a.empty()) {
+            effective = &a;
+        }
+        Vector<Uint32> b;
+        if (ShaderCompiler::LowerRectImages(*effective, b, false) && !b.empty()) effective = &b;
+        Vector<Uint32> c;
+        if (ShaderCompiler::Lower1DArrayImagesForEssl(*effective, c, false) && !c.empty()) effective = &c;
+        Vector<Uint32> d;
+        if (ShaderCompiler::LegalizeFragmentOutputIndexingForEssl(*effective, d, false) && !d.empty()) {
+            effective = &d;
+        }
+
+        SpvcSession session(*effective, SessionUsageBit::Transpile);
+        spvc_compiler_options options;
+        if (session.CreateOptions(&options) != SPVC_SUCCESS) return false;
+        spvc_compiler_options_set_uint(options, SPVC_COMPILER_OPTION_GLSL_VERSION, esslVersion);
+        spvc_compiler_options_set_bool(options, SPVC_COMPILER_OPTION_GLSL_ES, SPVC_TRUE);
+        spvc_compiler_options_set_bool(options, SPVC_COMPILER_OPTION_GLSL_VULKAN_SEMANTICS, SPVC_FALSE);
+        session.SetOptions(options);
+        const char* result = nullptr;
+        session.Compile(&result);
+        if (!result) return false;
+        outEssl = result;
+        return true;
+    }
+
+    EsslTranslationKeyInputs EsslInputsFor(const Vector<Uint32>& spirv) {
+        EsslTranslationKeyInputs inputs;
+        inputs.spirv = &spirv;
+        inputs.shaderType = GL_FRAGMENT_SHADER;
+        inputs.maxColorTextureSamples = 4;
+        inputs.maxIntegerSamples = 1;
+        inputs.maxDepthTextureSamples = 4;
+        inputs.advertisedMaxSamples = 4;
+        inputs.esslVersion = 320;
+        return inputs;
+    }
+} // namespace
+
+// ---------------------------------------------------------------------------------------
+// L1, in situ: the full glCompileShader + glLinkProgram path for a repeated program.
+// ---------------------------------------------------------------------------------------
+// Arg(0) = the CTS smoke size; Arg(120) = a heavy stage, bracketing the ratio.
+static void BM_ProgramLink_CacheOff(benchmark::State& state) {
+    MobileGL::Initialize();
+    const SyncCompileScope sync;
+    const CacheModeScope cache(false);
+    const String vs = kVertexSource;
+    const String fs = SwizzleLikeFragment("", static_cast<int>(state.range(0)));
+    for (auto _ : state) {
+        LinkOneProgram(vs, fs);
+    }
+    state.SetLabel("MOBILEGL_SHADER_CACHE=0");
+}
+BENCHMARK(BM_ProgramLink_CacheOff)->Arg(0)->Arg(120)->Unit(benchmark::kMicrosecond);
+
+static void BM_ProgramLink_CacheOn(benchmark::State& state) {
+    MobileGL::Initialize();
+    const SyncCompileScope sync;
+    const CacheModeScope cache(true);
+    const String vs = kVertexSource;
+    const String fs = SwizzleLikeFragment("", static_cast<int>(state.range(0)));
+    LinkOneProgram(vs, fs); // prime, so the measured loop is the steady state
+    const TranslationCacheStats before = MG_State::GLState::GetProgramTranslationCache().Stats();
+    const TranslationCacheStats parseBefore = GetShaderParseVerdictCache().Stats();
+    for (auto _ : state) {
+        LinkOneProgram(vs, fs);
+    }
+    const TranslationCacheStats stats = MG_State::GLState::GetProgramTranslationCache().Stats();
+    const TranslationCacheStats parseStats = GetShaderParseVerdictCache().Stats();
+    state.counters["L1_hits"] = static_cast<double>(stats.hits - before.hits);
+    state.counters["L1_misses"] = static_cast<double>(stats.misses - before.misses);
+    // Two stages per iteration, so a clean run shows L1c_hits == 2 * iterations and zero
+    // misses: every glCompileShader in the loop skipped its parse.
+    state.counters["L1c_hits"] = static_cast<double>(parseStats.hits - parseBefore.hits);
+    state.counters["L1c_misses"] = static_cast<double>(parseStats.misses - parseBefore.misses);
+}
+BENCHMARK(BM_ProgramLink_CacheOn)->Arg(0)->Arg(120)->Unit(benchmark::kMicrosecond);
+
+// ---------------------------------------------------------------------------------------
+// L1, the shape the memo actually exists for: MANY PROGRAMS OUT OF THE SAME SHADERS.
+//
+// The pair above deletes its shader objects every iteration, which forces a fresh glslang
+// parse per iteration no matter what the link does - glCompileShader parses, and that is a
+// DIFFERENT entry point from the one L1 memoizes. It is a real workload (what an application
+// that never reuses a shader object pays) but it is the pessimistic one, and the residual it
+// leaves is the parse, not the link.
+//
+// This pair keeps the shader objects alive, so the parses happen once before the measured
+// loop and the L1 hit then skips the link, mapIO, the SPIR-V, the reflection and the routing
+// outright.
+//
+// SINCE L1c THIS IS THE CONTROL, NOT THE TARGET. Nothing inside the measured loop calls
+// glCompileShader, so L1c cannot fire here at all - which is exactly what makes the pair
+// useful: it is the shape that says whether the compile-side memo has slowed the LINK path
+// down. Its numbers should be indistinguishable from the pre-L1c ones.
+// ---------------------------------------------------------------------------------------
+namespace {
+    struct SharedShaders {
+        GLuint vs = 0;
+        GLuint fs = 0;
+    };
+
+    SharedShaders MakeSharedShaders(const String& vertexSource, const String& fragmentSource) {
+        using namespace MG_Impl::GLImpl;
+        SharedShaders shaders;
+        shaders.vs = CreateShader(GL_VERTEX_SHADER);
+        const char* vsText = vertexSource.c_str();
+        ShaderSource(shaders.vs, 1, &vsText, nullptr);
+        CompileShader(shaders.vs);
+        shaders.fs = CreateShader(GL_FRAGMENT_SHADER);
+        const char* fsText = fragmentSource.c_str();
+        ShaderSource(shaders.fs, 1, &fsText, nullptr);
+        CompileShader(shaders.fs);
+        return shaders;
+    }
+
+    void LinkFromSharedShaders(const SharedShaders& shaders) {
+        using namespace MG_Impl::GLImpl;
+        const GLuint program = CreateProgram();
+        AttachShader(program, shaders.vs);
+        AttachShader(program, shaders.fs);
+        LinkProgram(program);
+        benchmark::DoNotOptimize(program);
+        DeleteProgram(program);
+    }
+} // namespace
+
+static void BM_SharedShaderLink_CacheOff(benchmark::State& state) {
+    MobileGL::Initialize();
+    const SyncCompileScope sync;
+    const CacheModeScope cache(false);
+    const SharedShaders shaders =
+        MakeSharedShaders(kVertexSource, SwizzleLikeFragment("", static_cast<int>(state.range(0))));
+    for (auto _ : state) {
+        LinkFromSharedShaders(shaders);
+    }
+    state.SetLabel("MOBILEGL_SHADER_CACHE=0");
+}
+BENCHMARK(BM_SharedShaderLink_CacheOff)->Arg(0)->Arg(120)->Unit(benchmark::kMicrosecond);
+
+static void BM_SharedShaderLink_CacheOn(benchmark::State& state) {
+    MobileGL::Initialize();
+    const SyncCompileScope sync;
+    const CacheModeScope cache(true);
+    const SharedShaders shaders =
+        MakeSharedShaders(kVertexSource, SwizzleLikeFragment("", static_cast<int>(state.range(0))));
+    LinkFromSharedShaders(shaders); // prime, so the measured loop is the steady state
+    const TranslationCacheStats before = MG_State::GLState::GetProgramTranslationCache().Stats();
+    for (auto _ : state) {
+        LinkFromSharedShaders(shaders);
+    }
+    const TranslationCacheStats stats = MG_State::GLState::GetProgramTranslationCache().Stats();
+    state.counters["L1_hits"] = static_cast<double>(stats.hits - before.hits);
+    state.counters["L1_misses"] = static_cast<double>(stats.misses - before.misses);
+}
+BENCHMARK(BM_SharedShaderLink_CacheOn)->Arg(0)->Arg(120)->Unit(benchmark::kMicrosecond);
+
+// ---------------------------------------------------------------------------------------
+// L2, component: the DirectGLES SPIR-V pass chain plus SPIRV-Cross for one stage.
+// ---------------------------------------------------------------------------------------
+static void BM_EsslTranspile_CacheOff(benchmark::State& state) {
+    MobileGL::Initialize();
+    const Vector<Uint32> spirv =
+        BuildSanitizedFragmentSpirv(SwizzleLikeFragment("", static_cast<int>(state.range(0))));
+    if (spirv.empty()) {
+        state.SkipWithError("could not build the fragment module");
+        return;
+    }
+    String essl;
+    for (auto _ : state) {
+        if (!TranspileLikeDirectGles(spirv, 320, essl)) {
+            state.SkipWithError("transpile failed");
+            break;
+        }
+        benchmark::DoNotOptimize(essl.data());
+    }
+    state.SetLabel("MOBILEGL_SHADER_CACHE=0");
+}
+BENCHMARK(BM_EsslTranspile_CacheOff)->Arg(0)->Arg(120)->Unit(benchmark::kMicrosecond);
+
+static void BM_EsslTranspile_CacheOn(benchmark::State& state) {
+    MobileGL::Initialize();
+    const Vector<Uint32> spirv =
+        BuildSanitizedFragmentSpirv(SwizzleLikeFragment("", static_cast<int>(state.range(0))));
+    if (spirv.empty()) {
+        state.SkipWithError("could not build the fragment module");
+        return;
+    }
+    BoundedTranslationCache<EsslTranslationResult> cache("bench L2", 64, 8u << 20);
+    const EsslTranslationKeyInputs inputs = EsslInputsFor(spirv);
+    for (auto _ : state) {
+        const TranslationCacheKey key = BuildEsslTranslationKey(inputs);
+        EsslTranslationResultPtr hit = cache.Find(key);
+        if (!hit) {
+            auto payload = MakeShared<EsslTranslationResult>();
+            if (!TranspileLikeDirectGles(spirv, inputs.esslVersion, payload->essl)) {
+                state.SkipWithError("transpile failed");
+                break;
+            }
+            cache.Insert(key, EsslTranslationResultPtr(payload), EsslTranslationResultBytes(*payload));
+            hit = payload;
+        }
+        benchmark::DoNotOptimize(hit->essl.data());
+    }
+    const TranslationCacheStats stats = cache.Stats();
+    state.counters["L2_hits"] = static_cast<double>(stats.hits);
+    state.counters["L2_misses"] = static_cast<double>(stats.misses);
+}
+BENCHMARK(BM_EsslTranspile_CacheOn)->Arg(0)->Arg(120)->Unit(benchmark::kMicrosecond);
+
+// ---------------------------------------------------------------------------------------
+// L1c, the shape where it could LOSE rather than win: the DEFERRED PARSE.
+// ---------------------------------------------------------------------------------------
+// A stage whose compile hits L1c holds no AST, so if the program-level key then MISSES, the
+// parse it skipped has to happen anyway - inside the link, via ClaimParsedShader. The parse
+// is moved, not removed, and this pair is what says whether moving it costs anything.
+//
+// The shape forces exactly that, every iteration: one CONSTANT vertex source (hits L1c after
+// the first iteration) linked against a FRESH fragment source each time (misses L1c, and
+// makes the program key miss too). So:
+//
+//   cache off - two parses at glCompileShader, then the link.
+//   cache on  - one parse at glCompileShader (the fragment), one deferred parse inside the
+//               link (the vertex), then the link.
+//
+// The parse count is identical, so these two should land within noise of each other. If the
+// On arm is materially SLOWER, L1c is charging for something - the per-compile key build and
+// hash over the full preprocessed source, or the loss of the claim-CAS reuse - and that cost
+// shows up here and nowhere else.
+//
+// The distinct fragment sources also churn both front-end levels through their FIFO caps,
+// which is the eviction behaviour a real shaderpack load produces; over a long run the
+// constant vertex entry is occasionally evicted by that churn and re-inserted, so the L1c
+// hit rate reported below is high but not exactly 1.0 per iteration.
+namespace {
+    String UniqueFragmentSource(const Uint64 serial, const int padLines) {
+        return SwizzleLikeFragment("", padLines) +
+               "\n// unique-" + std::to_string(serial) + "\n";
+    }
+} // namespace
+
+static void BM_DeferredParseLink_CacheOff(benchmark::State& state) {
+    MobileGL::Initialize();
+    const SyncCompileScope sync;
+    const CacheModeScope cache(false);
+    const String vs = kVertexSource;
+    Uint64 serial = 0;
+    for (auto _ : state) {
+        LinkOneProgram(vs, UniqueFragmentSource(serial++, static_cast<int>(state.range(0))));
+    }
+    state.SetLabel("MOBILEGL_SHADER_CACHE=0");
+}
+BENCHMARK(BM_DeferredParseLink_CacheOff)->Arg(0)->Arg(120)->Unit(benchmark::kMicrosecond);
+
+static void BM_DeferredParseLink_CacheOn(benchmark::State& state) {
+    MobileGL::Initialize();
+    const SyncCompileScope sync;
+    const CacheModeScope cache(true);
+    const String vs = kVertexSource;
+    Uint64 serial = 0;
+    LinkOneProgram(vs, UniqueFragmentSource(~0ull, static_cast<int>(state.range(0)))); // prime the vertex entry
+    const TranslationCacheStats before = MG_State::GLState::GetProgramTranslationCache().Stats();
+    const TranslationCacheStats parseBefore = GetShaderParseVerdictCache().Stats();
+    for (auto _ : state) {
+        LinkOneProgram(vs, UniqueFragmentSource(serial++, static_cast<int>(state.range(0))));
+    }
+    const TranslationCacheStats stats = MG_State::GLState::GetProgramTranslationCache().Stats();
+    const TranslationCacheStats parseStats = GetShaderParseVerdictCache().Stats();
+    // Expected shape: L1 all misses (every program is new), L1c one hit (vertex) and one miss
+    // (fragment) per iteration.
+    state.counters["L1_hits"] = static_cast<double>(stats.hits - before.hits);
+    state.counters["L1_misses"] = static_cast<double>(stats.misses - before.misses);
+    state.counters["L1c_hits"] = static_cast<double>(parseStats.hits - parseBefore.hits);
+    state.counters["L1c_misses"] = static_cast<double>(parseStats.misses - parseBefore.misses);
+}
+BENCHMARK(BM_DeferredParseLink_CacheOn)->Arg(0)->Arg(120)->Unit(benchmark::kMicrosecond);
+
+BENCHMARK_MAIN();

MobileGL/MG_Benchmark/Transpile/CMakeLists.txt

@@ -0,0 +1,20 @@
+cmake_minimum_required(VERSION 3.24)
+
+# Deliberately NOT a google-benchmark target: the interesting quantity is a per-stage
+# breakdown of one program build, which needs its own clock around sub-steps that share
+# set-up, and a plain main() keeps the output a table this can be read straight out of.
+add_executable(
+        TranspileProfile
+        TranspileProfile.cpp
+)
+
+target_include_directories(TranspileProfile PRIVATE
+        ${MGL_ROOT}/include
+        ${MGL_ROOT}/MobileGL
+        ${MGL_ROOT}/3rdparty/SPIRV-Reflect
+)
+
+target_link_libraries(
+        TranspileProfile PRIVATE
+        ${LINK_LIBRARIES}
+)

MobileGL/MG_Impl/CGLImpl/CGLImpl.cpp

@@ -24,6 +24,7 @@ namespace MobileGL::MG_Impl::CGLImpl {
             GLint Samples = 0;
             GLint Profile = kCGLOGLPVersion_3_2_Core;
             GLint RendererId = 0x4d474c;
+            GLint DisplayMask = 0;
         };
 
         struct ContextObject {
@@ -134,6 +135,9 @@ namespace MobileGL::MG_Impl::CGLImpl {
             case kCGLPFARendererID:
                 pixelFormat.RendererId = value;
                 break;
+            case kCGLPFADisplayMask:
+                pixelFormat.DisplayMask = value;
+                break;
             default:
                 break;
             }
@@ -343,6 +347,9 @@ namespace MobileGL::MG_Impl::CGLImpl {
         case kCGLPFARendererID:
             *value = pixelFormat->RendererId;
             return kCGLNoError;
+        case kCGLPFADisplayMask:
+            *value = pixelFormat->DisplayMask;
+            return kCGLNoError;
         case kCGLPFAOpenGLProfile:
             *value = pixelFormat->Profile;
             return kCGLNoError;
@@ -481,6 +488,32 @@ namespace MobileGL::MG_Impl::CGLImpl {
         return it == currentContexts.end() ? nullptr : it->second;
     }
 
+    CGLError SetVirtualScreen(CGLContextObj ctx, GLint screen) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        if (screen != 0) {
+            return kCGLBadValue;
+        }
+        object->VirtualScreen = screen;
+        return kCGLNoError;
+    }
+
+    CGLError GetVirtualScreen(CGLContextObj ctx, GLint* screen) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        if (!screen) {
+            return kCGLBadAddress;
+        }
+        *screen = object->VirtualScreen;
+        return kCGLNoError;
+    }
+
     CGLError SetParameter(CGLContextObj ctx, CGLContextParameter pname, const GLint* params) {
         const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
         auto* object = TryGetContext(ctx);

MobileGL/MG_Impl/CGLImpl/CGLImpl.h

@@ -32,6 +32,8 @@ namespace MobileGL::MG_Impl::CGLImpl {
 
     CGLError SetCurrentContext(CGLContextObj ctx);
     CGLContextObj GetCurrentContext();
+    CGLError SetVirtualScreen(CGLContextObj ctx, GLint screen);
+    CGLError GetVirtualScreen(CGLContextObj ctx, GLint* screen);
     CGLError SetParameter(CGLContextObj ctx, CGLContextParameter pname, const GLint* params);
     CGLError GetParameter(CGLContextObj ctx, CGLContextParameter pname, GLint* params);
     CGLError UpdateContext(CGLContextObj ctx);

MobileGL/MG_Impl/CGLImpl/Exporting/Definitions.cpp

@@ -71,6 +71,14 @@ MOBILEGL_CGL_API CGLContextObj CGLGetCurrentContext(void) {
     return MobileGL::MG_Impl::CGLImpl::GetCurrentContext();
 }
 
+MOBILEGL_CGL_API CGLError CGLSetVirtualScreen(CGLContextObj ctx, GLint screen) {
+    return MobileGL::MG_Impl::CGLImpl::SetVirtualScreen(ctx, screen);
+}
+
+MOBILEGL_CGL_API CGLError CGLGetVirtualScreen(CGLContextObj ctx, GLint* screen) {
+    return MobileGL::MG_Impl::CGLImpl::GetVirtualScreen(ctx, screen);
+}
+
 MOBILEGL_CGL_API CGLError CGLSetParameter(CGLContextObj ctx, CGLContextParameter pname, const GLint* params) {
     return MobileGL::MG_Impl::CGLImpl::SetParameter(ctx, pname, params);
 }

MobileGL/MG_Impl/DyldInterpose/DyldInterpose.cpp

@@ -10,8 +10,12 @@
 
 #if defined(__APPLE__)
 
+#include "MG_Impl/CGLImpl/CGLImpl.h"
 #include "MG_Impl/GetProcAddress.h"
 
+#include <CoreGraphics/CoreGraphics.h>
+#include <CoreVideo/CVDisplayLink.h>
+#include <cstdint>
 #include <dlfcn.h>
 
 namespace {
@@ -47,10 +51,52 @@ namespace {
         return dlsym(handle, symbol);
     }
 
+    CGDirectDisplayID DisplayForMask(GLint displayMask) {
+        constexpr std::uint32_t MaxDisplays = sizeof(CGOpenGLDisplayMask) * 8;
+        CGDirectDisplayID displays[MaxDisplays] = {};
+        std::uint32_t displayCount = 0;
+        if (displayMask != 0 &&
+            CGGetActiveDisplayList(MaxDisplays, displays, &displayCount) == kCGErrorSuccess) {
+            const auto mask = static_cast<CGOpenGLDisplayMask>(displayMask);
+            for (std::uint32_t i = 0; i < displayCount; ++i) {
+                if ((CGDisplayIDToOpenGLDisplayMask(displays[i]) & mask) != 0) {
+                    return displays[i];
+                }
+            }
+        }
+        return CGMainDisplayID();
+    }
+
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wdeprecated-declarations"
+    CVReturn MobileGLCVDisplayLinkSetCurrentCGDisplayFromOpenGLContext(
+        CVDisplayLinkRef displayLink,
+        CGLContextObj context,
+        CGLPixelFormatObj pixelFormat) {
+        GLint virtualScreen = 0;
+        if (MobileGL::MG_Impl::CGLImpl::GetVirtualScreen(context, &virtualScreen) == kCGLNoError) {
+            GLint displayMask = 0;
+            if (!displayLink ||
+                MobileGL::MG_Impl::CGLImpl::DescribePixelFormat(
+                    pixelFormat, virtualScreen, kCGLPFADisplayMask, &displayMask) != kCGLNoError) {
+                return kCVReturnInvalidArgument;
+            }
+            return CVDisplayLinkSetCurrentCGDisplay(displayLink, DisplayForMask(displayMask));
+        }
+
+        using OriginalFunction = CVReturn (*)(CVDisplayLinkRef, CGLContextObj, CGLPixelFormatObj);
+        static const auto original = reinterpret_cast<OriginalFunction>(
+            dlsym(RTLD_NEXT, "CVDisplayLinkSetCurrentCGDisplayFromOpenGLContext"));
+        return original ? original(displayLink, context, pixelFormat) : kCVReturnError;
+    }
+
     __attribute__((used)) static const DyldInterposeEntry kMobileGLDyldInterpose[]
         __attribute__((section("__DATA,__interpose"))) = {
             {reinterpret_cast<const void*>(MobileGLDlsym), reinterpret_cast<const void*>(dlsym)},
+            {reinterpret_cast<const void*>(MobileGLCVDisplayLinkSetCurrentCGDisplayFromOpenGLContext),
+             reinterpret_cast<const void*>(CVDisplayLinkSetCurrentCGDisplayFromOpenGLContext)},
         };
+#pragma clang diagnostic pop
 } // namespace
 
 #endif

MobileGL/MG_Impl/DyldInterpose/ExportedSymbols.txt

@@ -0,0 +1,10 @@
+# Public CGL entry points.
+_CGL*
+
+# Public EGL entry points.
+_egl*
+
+# Public OpenGL and GLX entry points. OpenGL function names always use an
+# uppercase letter or digit after the "gl" prefix; excluding lowercase here
+# deliberately prevents glslang_* from matching this pattern.
+_gl[A-Z0-9]*

MobileGL/MG_Impl/EGLImpl/EGLImpl.cpp

@@ -8,6 +8,7 @@
 
 #include "EGLImpl.h"
 #include "../GetProcAddress.h"
+#include <Init.h>
 #include <MG_Backend/BackendObjects.h>
 #include <MG_State/EGLState/Core.h>
 #include <mutex>
@@ -20,11 +21,22 @@ 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();
         }
 
+        // Entry points that can legitimately be an application's FIRST EGL
+        // call (display/proc-address/string queries) lazily bring MobileGL
+        // up here, so the library needs no static constructor and can
+        // re-initialize after the last eglTerminate tore everything down.
+        // Teardown-ish entry points keep using GetState() and fail benignly
+        // when MobileGL is not initialized.
+        EGLStateContext* GetStateEnsureInitialized() {
+            MobileGL::EnsureInitialized();
+            return GetState();
+        }
+
         MG_Backend::BackendObject* GetBackendObject(EGLStateContext* state) {
             auto* backendObject = MG_Backend::pActiveBackendObject.get();
             if (!backendObject && state) {
@@ -49,6 +61,8 @@ namespace MobileGL::MG_Impl::EGLImpl {
             return MG_Backend::WindowBackend::Android;
 #elif defined(__APPLE__)
             return MG_Backend::WindowBackend::MetalLayer;
+#elif defined(_WIN32)
+            return MG_Backend::WindowBackend::Win32;
 #elif defined(__linux__)
             return MG_Backend::WindowBackend::X11;
 #else
@@ -132,7 +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;
         }
@@ -158,11 +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;
         }
@@ -187,7 +201,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     EGLBoolean Initialize(EGLDisplay dpy, EGLint* major, EGLint* minor) {
-        auto* state = GetState();
+        auto* state = GetStateEnsureInitialized();
         if (!state) {
             return EGL_FALSE;
         }
@@ -197,7 +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)) {
@@ -208,7 +222,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     EGLDisplay GetDisplay(NativeDisplayType display) {
-        auto* state = GetState();
+        auto* state = GetStateEnsureInitialized();
         if (!state) {
             return EGL_NO_DISPLAY;
         }
@@ -251,7 +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;
@@ -263,12 +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);
@@ -313,6 +327,14 @@ namespace MobileGL::MG_Impl::EGLImpl {
         if (auto* backendObject = MG_Backend::pActiveBackendObject.get()) {
             backendObject->ReleaseEGLResources();
         }
+        // The last initialized display is gone and nothing is current on any
+        // thread: tear the whole library down deterministically inside the
+        // EGL lifecycle (backend, GL/EGL state, glslang). A later EGL call
+        // re-initializes lazily via GetStateEnsureInitialized(); process exit
+        // then has nothing left to destroy.
+        if (!state->HasAnyInitializedDisplay() && !state->HasAnyCurrentContext()) {
+            MobileGL::Destroy();
+        }
         return EGL_TRUE;
     }
 
@@ -345,7 +367,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     EGLBoolean BindAPI(EGLenum api) {
-        auto* state = GetState();
+        auto* state = GetStateEnsureInitialized();
         if (!state) {
             return EGL_FALSE;
         }
@@ -378,7 +400,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     char const* QueryString(EGLDisplay display, EGLint name) {
-        auto* state = GetState();
+        auto* state = GetStateEnsureInitialized();
         if (!state) {
             return nullptr;
         }
@@ -641,7 +663,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     EGLDisplay GetPlatformDisplay(EGLenum platform, void* native_display, const EGLAttrib* attrib_list) {
         (void)attrib_list;
 
-        auto* state = GetState();
+        auto* state = GetStateEnsureInitialized();
         if (!state) {
             return EGL_NO_DISPLAY;
         }
@@ -681,7 +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;
         }
@@ -704,7 +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);
@@ -737,11 +759,12 @@ namespace MobileGL::MG_Impl::EGLImpl {
         if (!name) {
             return nullptr;
         }
+        MobileGL::EnsureInitialized();
 
         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

@@ -8,12 +8,17 @@
 
 #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>
 #include <MG_State/GLState/Core.h>
 #include <MG_State/GLState/ErrorState/Error.h>
 #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 {
@@ -27,6 +32,8 @@ namespace MobileGL::MG_Impl::GLImpl {
             NamedBufferData,
             NamedBufferSubData,
             CopyNamedBufferSubData,
+            ClearBufferData,
+            ClearBufferSubData,
             ClearNamedBufferData,
             ClearNamedBufferSubData,
             MapBufferRange,
@@ -38,6 +45,7 @@ namespace MobileGL::MG_Impl::GLImpl {
             GetNamedBufferParameteriv,
             GetNamedBufferParameteri64v,
             GetNamedBufferPointerv,
+            GetNamedBufferSubData,
         };
 
         const char* GetBufferOpName(BufferOp op) {
@@ -60,6 +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:
@@ -76,6 +88,8 @@ namespace MobileGL::MG_Impl::GLImpl {
                 return "UnmapNamedBuffer";
             case BufferOp::FlushMappedNamedBufferRange:
                 return "FlushMappedNamedBufferRange";
+            case BufferOp::GetNamedBufferSubData:
+                return "GetNamedBufferSubData";
             case BufferOp::GetNamedBufferParameteriv:
                 return "GetNamedBufferParameteriv";
             case BufferOp::GetNamedBufferParameteri64v:
@@ -89,25 +103,54 @@ namespace MobileGL::MG_Impl::GLImpl {
 
         SharedPtr<MG_State::GLState::BufferObject> GetNamedBufferObject(GLuint buffer, BufferOp op);
 
+        // The size of one cleared element, which is what offset and size must be multiples of
+        // (GL 4.6 core 6.3). `internalformat` is restricted to the buffer-texture format table, and
+        // `format`/`type` describe the client-side pattern, so both are validated here and the
+        // caller only has to know how wide an element is.
         SizeT GetClearPatternSize(GLenum internalformat, GLenum format, GLenum type, BufferOp op) {
-            if (format != GL_RED_INTEGER) {
+            if (!IsBufferTextureInternalFormat(internalformat)) {
                 MG_State::pGLContext->RecordError(
                     ErrorCode::InvalidEnum,
-                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", GetBufferOpName(op),
-                                                 "Only GL_RED_INTEGER buffer clears are currently supported."));
+                    MakeUnique<GenericErrorInfo>(
+                        "MG_Impl/GLImpl", GetBufferOpName(op),
+                        std::format("internalformat 0x{:X} is not one of the sized formats a buffer clear accepts.",
+                                    internalformat)));
                 return 0;
             }
 
-            if (internalformat == GL_R8UI && type == GL_UNSIGNED_BYTE) return sizeof(GLubyte);
-            if (internalformat == GL_R32UI && type == GL_UNSIGNED_INT) return sizeof(GLuint);
+            // Unlike internalformat, a bad format or type here is INVALID_VALUE rather than
+            // INVALID_ENUM (GL 4.6 core 6.3) - the odd one out among the enum arguments.
+            const TextureInputFormat inputFormat = MG_Util::ConvertGLEnumToTextureInputFormat(format);
+            if (inputFormat == TextureInputFormat::Unknown) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", GetBufferOpName(op),
+                                                 std::format("format 0x{:X} is not a pixel format.", format)));
+                return 0;
+            }
 
-            MG_State::pGLContext->RecordError(
-                ErrorCode::InvalidEnum,
-                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", GetBufferOpName(op),
-                                             std::format("Unsupported clear format tuple: internalformat=0x{:X}, "
-                                                         "format=0x{:X}, type=0x{:X}",
-                                                         internalformat, format, type)));
-            return 0;
+            const TexturePixelDataType pixelType = MG_Util::ConvertGLEnumToTexturePixelDataType(type);
+            if (pixelType == TexturePixelDataType::Unknown) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", GetBufferOpName(op),
+                                                 std::format("type 0x{:X} is not a pixel type.", type)));
+                return 0;
+            }
+
+            const TextureInternalFormat internal =
+                MG_Util::ConvertGLEnumToTextureInternalFormat(internalformat);
+            const SizeT elementSize = MG_Util::GetSizedInternalFormatSizeInBytes(internal);
+            if (elementSize == 0) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidEnum,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", GetBufferOpName(op),
+                                                 std::format("internalformat 0x{:X} has no known element size.",
+                                                             internalformat)));
+                return 0;
+            }
+
+            return elementSize;
         }
 
         Bool ValidateBufferClearRange(const SharedPtr<MG_State::GLState::BufferObject>& bufferObject, GLintptr offset,
@@ -148,27 +191,59 @@ namespace MobileGL::MG_Impl::GLImpl {
             return true;
         }
 
-        void ClearNamedBufferRange_State(GLuint buffer, GLenum internalformat, GLintptr offset, GLsizeiptr size,
-                                         GLenum format, GLenum type, const void* data, BufferOp op) {
+        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));
-            if (data) {
-                const auto* pattern = static_cast<const Uint8*>(data);
-                for (SizeT at = 0; at < clearData.size(); at += patternSize) {
-                    Memcpy(clearData.data() + at, pattern, patternSize);
-                }
-            } else {
-                Memset(clearData.data(), 0, clearData.size());
-            }
-
-            bufferObject->UploadSubData({clearData.data(), clearData.size()}, static_cast<SizeT>(offset));
+            Vector<Uint8> pattern;
+            if (!BuildClearPattern(internalformat, format, type, data, patternSize, op, pattern)) return;
+            bufferObject->FillSubData({pattern.data(), pattern.size()}, static_cast<SizeT>(offset),
+                                      static_cast<SizeT>(size));
         }
 
         auto& GetBufferBindingSlot(BufferTarget target) {
@@ -330,12 +405,21 @@ namespace MobileGL::MG_Impl::GLImpl {
                     } else if (access & BufferMappingAccessBit::Write) {
                         *params = GL_WRITE_ONLY;
                     } else {
-                        *params = 0;
+                        *params = GL_READ_WRITE;
                     }
                 } else {
-                    *params = 0;
+                    // Initial value, and what glUnmapBuffer restores (GL 4.6 core table 6.2).
+                    *params = GL_READ_WRITE;
                 }
                 break;
+            case GL_BUFFER_ACCESS_FLAGS:
+                // The MapBufferRange flags verbatim; glMapBuffer's access enum has already been
+                // normalised into the same bits. Zero while the buffer is not mapped.
+                *params = bufferObject->IsMapped()
+                    ? static_cast<GLint>(
+                          MG_Util::ConvertBufferMappingAccessToGLEnum(bufferObject->GetMappingAccess()))
+                    : 0;
+                break;
             case GL_BUFFER_MAPPED:
                 *params = bufferObject->IsMapped() ? GL_TRUE : GL_FALSE;
                 break;
@@ -807,6 +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(
@@ -817,18 +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);
     }
 
@@ -878,6 +954,45 @@ namespace MobileGL::MG_Impl::GLImpl {
             return;
         }
 
+        bufferObject->SyncGpuWrites();
+        bufferObject->DownloadSubData(data, static_cast<SizeT>(offset), static_cast<SizeT>(size));
+    }
+
+    void GetNamedBufferSubData_State(GLuint buffer, GLintptr offset, GLsizeiptr size, void* data) {
+        if (!data) {
+            // Match GetBufferSubData_State: a null pointer is a caller bug, not a GL-specified error.
+            return;
+        }
+
+        if (size < 0 || offset < 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "GetNamedBufferSubData_State",
+                                             "Offset and size must be non-negative."));
+            return;
+        }
+
+        auto bufferObject = GetNamedBufferObject(buffer, BufferOp::GetNamedBufferSubData);
+        if (!bufferObject) return;
+
+        if (static_cast<SizeT>(offset) + static_cast<SizeT>(size) > bufferObject->GetSize()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "GetNamedBufferSubData_State",
+                                             "Offset and size exceed buffer size."));
+            return;
+        }
+
+        if (bufferObject->IsMapped() &&
+            !(bufferObject->GetMappingAccess() & BufferMappingAccessBit::Persistent)) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "GetNamedBufferSubData_State",
+                                             "Cannot read from a buffer object mapped without GL_MAP_PERSISTENT_BIT."));
+            return;
+        }
+
+        bufferObject->SyncGpuWrites();
         bufferObject->DownloadSubData(data, static_cast<SizeT>(offset), static_cast<SizeT>(size));
     }
 
@@ -920,6 +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,
@@ -928,8 +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,
@@ -964,6 +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,
@@ -972,8 +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,
@@ -1106,17 +1226,34 @@ 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,
-                                    type, data, BufferOp::ClearNamedBufferData);
+        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,
-                                    BufferOp::ClearNamedBufferSubData);
+        auto bufferObject = GetNamedBufferObject(buffer, BufferOp::ClearNamedBufferSubData);
+        if (!bufferObject) return;
+        ClearBufferRange_State(bufferObject, internalformat, offset, size, format, type, data,
+                               BufferOp::ClearNamedBufferSubData);
     }
 
     void* MapNamedBuffer_State(GLuint buffer, GLenum access) {
@@ -1351,6 +1488,14 @@ namespace MobileGL::MG_Impl::GLImpl {
         BufferTarget bufferTarget = MG_Util::ConvertGLEnumToBufferTarget(target);
         if (!BufferImpl::ValidateBufferBindingPointTarget(bufferTarget)) return;
         if (!BufferImpl::ValidateBufferBindingPointIndex(bufferTarget, pointIndex)) return;
+        if (bufferTarget == BufferTarget::TransformFeedback && MG_State::pGLContext->IsTransformFeedbackActive()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Transform feedback buffer bindings cannot change while transform "
+                                             "feedback is active."));
+            return;
+        }
         MG_State::pGLContext->TouchBufferBindingPoint(bufferTarget, pointIndex);
 
         auto& point = MG_State::pGLContext->GetBufferBindingPoint(bufferTarget, pointIndex);
@@ -1358,6 +1503,7 @@ namespace MobileGL::MG_Impl::GLImpl {
         if (buffer == 0) {
             point.Bind(nullptr);
             point.SetRange(Range1D(0, 0));
+            GetBufferBindingSlot(bufferTarget).Bind(nullptr);
             return;
         }
 
@@ -1376,6 +1522,81 @@ namespace MobileGL::MG_Impl::GLImpl {
         } else {
             point.ClearRange();
         }
+        // The indexed bind also binds to the generic binding point of the same target
+        // (GL 4.6 core 6.1.1). Callers rely on it: the texture_gather tests set up their
+        // SSBO with BindBufferBase and then size it through glBufferData on the generic
+        // target alone, which would otherwise raise GL_INVALID_OPERATION and leave the
+        // buffer with no storage.
+        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) {
@@ -1384,6 +1605,20 @@ namespace MobileGL::MG_Impl::GLImpl {
         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,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Transform feedback buffer bindings cannot change while transform "
+                                             "feedback is active."));
+            return;
+        }
         MG_State::pGLContext->TouchBufferBindingPoint(bufferTarget, index);
 
         auto& point = MG_State::pGLContext->GetBufferBindingPoint(bufferTarget, index);
@@ -1391,6 +1626,7 @@ namespace MobileGL::MG_Impl::GLImpl {
         if (buffer == 0) {
             point.Bind(nullptr);
             point.SetRange(Range1D(0, 0));
+            GetBufferBindingSlot(bufferTarget).Bind(nullptr);
             return;
         }
 
@@ -1408,6 +1644,8 @@ namespace MobileGL::MG_Impl::GLImpl {
         } else {
             point.ClearRange();
         }
+        // Also the generic binding point, exactly as BindBufferBase (GL 4.6 core 6.1.1).
+        GetBufferBindingSlot(bufferTarget).Bind(bufferObject);
     }
 
     /* @INSERTION_POINT:FUNCTION_IMPLEMENTATION@ */
@@ -1470,6 +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);
     }
@@ -1517,6 +1764,10 @@ namespace MobileGL::MG_Impl::GLImpl {
         BufferSubData_State(target, offset, size, data);
     }
 
+    void GetNamedBufferSubData(GLuint buffer, GLintptr offset, GLsizeiptr size, void* data) {
+        GetNamedBufferSubData_State(buffer, offset, size, data);
+    }
+
     void GetBufferSubData(GLenum target, GLintptr offset, GLsizeiptr size, void* data) {
         GetBufferSubData_State(target, offset, size, data);
     }
@@ -1542,16 +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);
@@ -41,6 +44,7 @@ namespace MobileGL::MG_Impl::GLImpl {
                            GLsizeiptr size);
     void BufferSubData(GLenum target, GLintptr offset, GLsizeiptr size, const void* data);
     void GetBufferSubData(GLenum target, GLintptr offset, GLsizeiptr size, void* data);
+    void GetNamedBufferSubData(GLuint buffer, GLintptr offset, GLsizeiptr size, void* data);
     void BufferData(GLenum target, GLsizeiptr size, const void* data, GLenum usage);
     void BindBuffer(GLenum target, GLuint buffer);
     void GenBuffers(GLsizei n, GLuint* buffers);

MobileGL/MG_Impl/GLImpl/Buffer/Validators.cpp

@@ -13,6 +13,7 @@
 #include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
 #include <MG_Util/Converters/MGToGL/BufferEnumConverter.h>
 #include <MG_Util/Converters/MGToStr/BufferEnumConverter.h>
+#include <MG_Util/ShaderTranspiler/Types.h>
 
 namespace MobileGL::MG_Impl::GLImpl::BufferImpl {
     Bool ValidateBufferTarget(BufferTarget target) {
@@ -53,13 +54,53 @@ namespace MobileGL::MG_Impl::GLImpl::BufferImpl {
         return true;
     }
 
-    Bool ValidateBufferBindingPointIndex(BufferTarget target, Uint index) {
-        SizeT pointCount = MG_State::pGLContext->GetBufferBindingPointCount(target);
-        if (target == BufferTarget::ShaderStorage && MG_Backend::pActiveBackendObject) {
-            const Int backendCount =
-                MG_Backend::pActiveBackendObject->GetDynamicParameters().MaxShaderStorageBufferBindings;
-            pointCount = std::min(pointCount, static_cast<SizeT>(std::max(backendCount, 0)));
+    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 =
+                    MG_Backend::pActiveBackendObject->GetDynamicParameters().MaxShaderStorageBufferBindings;
+                pointCount = std::min(pointCount, static_cast<SizeT>(std::max(backendCount, 0)));
+            }
+            if (target == BufferTarget::TransformFeedback) {
+                // GL_MAX_TRANSFORM_FEEDBACK_SEPARATE_ATTRIBS bounds the indexed capture
+                // binding points in GL 3.3 (no ARB_transform_feedback3).
+                pointCount = std::min<SizeT>(pointCount, 4);
+            }
+            if (target == BufferTarget::AtomicCounter) {
+                // GL_MAX_ATOMIC_COUNTER_BUFFER_BINDINGS, which is NOT the state layer's array
+                // size: a counter buffer reaches a shader only as a lowered storage block, so the
+                // reserved range is the ceiling, and glGetIntegerv advertises the same number.
+                pointCount = std::min<SizeT>(
+                    pointCount, static_cast<SizeT>(MG_Util::ShaderTranspiler::MAX_ATOMIC_COUNTER_BUFFER_BINDINGS));
+            }
+            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;
@@ -107,14 +148,10 @@ namespace MobileGL::MG_Impl::GLImpl::BufferImpl {
     }
 
     Bool ValidateBufferMappingAccess(Flags<BufferMappingAccessBit> accessBits) {
-        if (accessBits == BufferMappingAccessBit::Null) {
-            MG_State::pGLContext->RecordError(ErrorCode::InvalidEnum,
-                                              MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl/BufferImpl",
-                                                                           "ValidateBufferMappingAccess",
-                                                                           "Access bits cannot be null."));
-            return false;
-        }
-
+        // An empty mask is a legal value for a bitfield - it just fails the rule that a mapping
+        // must ask for read or write access, which is INVALID_OPERATION and belongs to the callers
+        // (both of them check it immediately after this). Rejecting it here as INVALID_ENUM
+        // reported the wrong error and hid theirs.
         const auto validBits = BufferMappingAccessBit::Read | BufferMappingAccessBit::Write |
                                BufferMappingAccessBit::InvalidateRange | BufferMappingAccessBit::InvalidateBuffer |
                                BufferMappingAccessBit::FlushExplicit | BufferMappingAccessBit::Unsynchronized |

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/Debug/GL_Debug.cpp

@@ -0,0 +1,271 @@
+// MobileGL - MobileGL/MG_Impl/GLImpl/Debug/GL_Debug.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 "GL_Debug.h"
+
+#include <cstring>
+
+#include <MG_State/GLState/Core.h>
+#include <MG_State/GLState/ErrorState/Error.h>
+#include <MG_Impl/GLImpl/Query/GL_Query.h>
+#include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
+
+namespace MobileGL::MG_Impl::GLImpl {
+    namespace {
+        // Must agree with what GL_Getter answers for GL_MAX_DEBUG_GROUP_STACK_DEPTH and
+        // GL_MAX_DEBUG_MESSAGE_LENGTH / GL_MAX_LABEL_LENGTH; an application that sizes a buffer
+        // off the query and then trips a different limit here would have no way to explain it.
+        constexpr SizeT kMaxDebugGroupStackDepth = 64;
+        constexpr GLsizei kMaxDebugMessageLength = 1024;
+        constexpr GLsizei kMaxLabelLength = 256;
+
+        // The debug state KHR_debug makes per-context. Held here rather than on GLContext because
+        // nothing else in MobileGL reads it, and it is keyed on the context id so a
+        // destroyed-and-recreated context starts with an empty stack and no labels - which the
+        // unit tests, which recreate the context between cases, depend on.
+        struct DebugState {
+            Uint64 contextId = 0;
+            // The messages pushed with glPushDebugGroup, innermost last. The base group GL creates
+            // the context with is implicit and is what makes the reported depth start at 1.
+            Vector<String> groupStack;
+            // Keyed by (identifier, name); see MakeObjectLabelKey.
+            UnorderedMap<Uint64, String> objectLabels;
+        };
+
+        DebugState& State() {
+            static DebugState state;
+            const Uint64 contextId = MG_State::pGLContext ? MG_State::pGLContext->GetTextureContextId() : 0;
+            if (state.contextId != contextId) {
+                state.contextId = contextId;
+                state.groupStack.clear();
+                state.objectLabels.clear();
+            }
+            return state;
+        }
+
+        Uint64 MakeObjectLabelKey(GLenum identifier, GLuint name) {
+            return (static_cast<Uint64>(identifier) << 32) | static_cast<Uint64>(name);
+        }
+
+        void RecordDebugError(ErrorCode code, const char* caller, const String& message) {
+            MG_State::pGLContext->RecordError(code, MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", caller, message));
+        }
+
+        // GL 4.6 core 20.2: only an APPLICATION or THIRD_PARTY source may be injected; the rest
+        // are reserved for the implementation itself.
+        Bool ValidateInjectedSource(GLenum source, const char* caller) {
+            if (source == GL_DEBUG_SOURCE_APPLICATION || source == GL_DEBUG_SOURCE_THIRD_PARTY) {
+                return true;
+            }
+            RecordDebugError(ErrorCode::InvalidEnum, caller,
+                             std::format("source {} is not GL_DEBUG_SOURCE_APPLICATION or "
+                                         "GL_DEBUG_SOURCE_THIRD_PARTY.",
+                                         MG_Util::ConvertGLEnumToString(source)));
+            return false;
+        }
+
+        // A negative length means the string is NUL-terminated (GL 4.6 core 20.2), which is how
+        // every one of these entry points spells "just use the whole thing".
+        Bool ValidateDebugStringLength(GLsizei length, const GLchar* text, GLsizei limit, const char* caller,
+                                       const char* what) {
+            const GLsizei effective =
+                length < 0 ? static_cast<GLsizei>(text != nullptr ? std::strlen(text) : 0) : length;
+            if (effective < limit) {
+                return true;
+            }
+            RecordDebugError(ErrorCode::InvalidValue, caller,
+                             std::format("{} length {} is not less than the {} limit of {}.", what, effective, what,
+                                         limit));
+            return false;
+        }
+
+        String MakeDebugString(GLsizei length, const GLchar* text) {
+            if (text == nullptr) return {};
+            return length < 0 ? String(text) : String(text, static_cast<SizeT>(length));
+        }
+
+        // Whether `name` currently names an object of `identifier`'s type. GL 4.6 core 20.5 makes
+        // labelling something that does not exist INVALID_VALUE, and every type KHR_debug lists
+        // has a frontend name check - so this is answered exactly rather than waved through.
+        // GL_DISPLAY_LIST is deliberately absent: it exists only in the compatibility profile,
+        // which MobileGL does not expose, so it falls to the INVALID_ENUM path below.
+        Bool ValidateLabelledObject(GLenum identifier, GLuint name, Bool& outIdentifierKnown) {
+            outIdentifierKnown = true;
+            auto* context = MG_State::pGLContext.get();
+            switch (identifier) {
+            case GL_BUFFER:
+                return context->ValidateBufferName(name);
+            case GL_SHADER:
+                return context->ValidateShaderName(name);
+            case GL_PROGRAM:
+                return context->ValidateProgramName(name);
+            case GL_VERTEX_ARRAY:
+                return context->ValidateVertexArrayName(name);
+            case GL_QUERY:
+                return IsQuery(name) == GL_TRUE;
+            case GL_PROGRAM_PIPELINE:
+                return context->ValidateProgramPipelineName(name);
+            case GL_TRANSFORM_FEEDBACK:
+                return context->ValidateTransformFeedbackName(name);
+            case GL_SAMPLER:
+                return context->ValidateSamplerName(name);
+            case GL_TEXTURE:
+                return context->ValidateTextureName(name);
+            case GL_RENDERBUFFER:
+                return context->ValidateRenderbufferName(name);
+            case GL_FRAMEBUFFER:
+                // Name 0 is the default framebuffer, which is a real, labellable object.
+                return name == 0 || context->ValidateFramebufferName(name);
+            default:
+                outIdentifierKnown = false;
+                return false;
+            }
+        }
+    } // namespace
+
+    GLint GetDebugGroupStackDepth() {
+        // GL 4.6 core 20.6: the context is created with one group already on the stack, so the
+        // reported depth is one more than the number of pushes the application has made.
+        return static_cast<GLint>(State().groupStack.size()) + 1;
+    }
+
+    void PushDebugGroup(GLenum source, GLuint id, GLsizei length, const GLchar* message) {
+        static_cast<void>(id);
+        if (!ValidateInjectedSource(source, __func__)) return;
+        if (!ValidateDebugStringLength(length, message, kMaxDebugMessageLength, __func__, "message")) return;
+
+        auto& state = State();
+        if (state.groupStack.size() + 1 >= kMaxDebugGroupStackDepth) {
+            // Not INVALID_*: KHR_debug gives the group stack its own error code.
+            RecordDebugError(ErrorCode::StackOverflow, __func__,
+                             std::format("the debug group stack is already {} deep, which is its maximum.",
+                                         kMaxDebugGroupStackDepth));
+            return;
+        }
+        state.groupStack.push_back(MakeDebugString(length, message));
+        MGLOG_D("glPushDebugGroup(%s) -> depth %d", state.groupStack.back().c_str(), GetDebugGroupStackDepth());
+    }
+
+    void PopDebugGroup() {
+        auto& state = State();
+        if (state.groupStack.empty()) {
+            // The base group the context was created with may not be popped (GL 4.6 core 20.6).
+            RecordDebugError(ErrorCode::StackUnderflow, __func__,
+                             "the debug group stack holds only the group the context was created with.");
+            return;
+        }
+        MGLOG_D("glPopDebugGroup(%s)", state.groupStack.back().c_str());
+        state.groupStack.pop_back();
+    }
+
+    void DebugMessageInsert(GLenum source, GLenum type, GLuint id, GLenum severity, GLsizei length,
+                            const GLchar* buf) {
+        static_cast<void>(id);
+        if (!ValidateInjectedSource(source, __func__)) return;
+        switch (type) {
+        case GL_DEBUG_TYPE_ERROR:
+        case GL_DEBUG_TYPE_DEPRECATED_BEHAVIOR:
+        case GL_DEBUG_TYPE_UNDEFINED_BEHAVIOR:
+        case GL_DEBUG_TYPE_PORTABILITY:
+        case GL_DEBUG_TYPE_PERFORMANCE:
+        case GL_DEBUG_TYPE_MARKER:
+        case GL_DEBUG_TYPE_PUSH_GROUP:
+        case GL_DEBUG_TYPE_POP_GROUP:
+        case GL_DEBUG_TYPE_OTHER:
+            break;
+        default:
+            RecordDebugError(ErrorCode::InvalidEnum, __func__,
+                             std::format("type {} is not a debug message type.",
+                                         MG_Util::ConvertGLEnumToString(type)));
+            return;
+        }
+        switch (severity) {
+        case GL_DEBUG_SEVERITY_HIGH:
+        case GL_DEBUG_SEVERITY_MEDIUM:
+        case GL_DEBUG_SEVERITY_LOW:
+        case GL_DEBUG_SEVERITY_NOTIFICATION:
+            break;
+        default:
+            RecordDebugError(ErrorCode::InvalidEnum, __func__,
+                             std::format("severity {} is not a debug message severity.",
+                                         MG_Util::ConvertGLEnumToString(severity)));
+            return;
+        }
+        if (!ValidateDebugStringLength(length, buf, kMaxDebugMessageLength, __func__, "message")) return;
+
+        // No callback is ever invoked and the message log is empty by construction
+        // (GL_MAX_DEBUG_LOGGED_MESSAGES is 1 and glGetDebugMessageLog returns nothing), so the
+        // application-visible effect is exactly the error checking above. The text still reaches
+        // MobileGL's own log, where it is worth having next to the calls it annotates - at debug
+        // level, so an application that inserts a message per draw costs nothing in a release build.
+        MGLOG_D("glDebugMessageInsert: %s", MakeDebugString(length, buf).c_str());
+    }
+
+    void ObjectLabel(GLenum identifier, GLuint name, GLsizei length, const GLchar* label) {
+        Bool identifierKnown = false;
+        const Bool objectExists = ValidateLabelledObject(identifier, name, identifierKnown);
+        if (!identifierKnown) {
+            RecordDebugError(ErrorCode::InvalidEnum, __func__,
+                             std::format("identifier {} is not a labellable object type.",
+                                         MG_Util::ConvertGLEnumToString(identifier)));
+            return;
+        }
+        if (!objectExists) {
+            RecordDebugError(ErrorCode::InvalidValue, __func__,
+                             std::format("{} {} is not the name of an existing object.",
+                                         MG_Util::ConvertGLEnumToString(identifier), name));
+            return;
+        }
+        if (!ValidateDebugStringLength(length, label, kMaxLabelLength, __func__, "label")) return;
+
+        auto& labels = State().objectLabels;
+        const Uint64 key = MakeObjectLabelKey(identifier, name);
+        if (label == nullptr) {
+            // GL 4.6 core 20.5: a NULL label removes any label the object had.
+            labels.erase(key);
+            return;
+        }
+        labels[key] = MakeDebugString(length, label);
+    }
+
+    void GetObjectLabel(GLenum identifier, GLuint name, GLsizei bufSize, GLsizei* length, GLchar* label) {
+        if (bufSize < 0) {
+            RecordDebugError(ErrorCode::InvalidValue, __func__, "bufSize must not be negative.");
+            return;
+        }
+        Bool identifierKnown = false;
+        const Bool objectExists = ValidateLabelledObject(identifier, name, identifierKnown);
+        if (!identifierKnown) {
+            RecordDebugError(ErrorCode::InvalidEnum, __func__,
+                             std::format("identifier {} is not a labellable object type.",
+                                         MG_Util::ConvertGLEnumToString(identifier)));
+            return;
+        }
+        if (!objectExists) {
+            RecordDebugError(ErrorCode::InvalidValue, __func__,
+                             std::format("{} {} is not the name of an existing object.",
+                                         MG_Util::ConvertGLEnumToString(identifier), name));
+            return;
+        }
+
+        const auto& labels = State().objectLabels;
+        const auto it = labels.find(MakeObjectLabelKey(identifier, name));
+        const String& text = it != labels.end() ? it->second : String{};
+        // GL 4.6 core 20.5: the returned length excludes the NUL, and an unlabelled object hands
+        // back an empty string with length 0 rather than an error.
+        SizeT copied = 0;
+        if (label != nullptr && bufSize > 0) {
+            copied = std::min(text.size(), static_cast<SizeT>(bufSize) - 1);
+            std::memcpy(label, text.data(), copied);
+            label[copied] = '\0';
+        }
+        if (length != nullptr) {
+            *length = static_cast<GLsizei>(copied);
+        }
+    }
+} // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Debug/GL_Debug.h

@@ -0,0 +1,42 @@
+// MobileGL - MobileGL/MG_Impl/GLImpl/Debug/GL_Debug.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_Impl::GLImpl {
+    // KHR_debug, core since GL 4.3 (GL 4.6 core 20). Applications use these to annotate a capture
+    // and to name their objects; Better Clouds calls all four for exactly that.
+    //
+    // MobileGL implements the STATE and the ERRORS, and deliberately does not forward the calls to
+    // the host driver. Two independent reasons:
+    //
+    //  * glObjectLabel names a FRONTEND object. MobileGL's texture 5 is not the ES driver's
+    //    texture 5 (and under DirectVulkan it is not a driver object at all), so forwarding the
+    //    pair verbatim would label an unrelated object or a nonexistent one - worse than not
+    //    labelling.
+    //  * A debug GROUP is only meaningful if it brackets the commands the application issued
+    //    inside it. Neither backend emits its work at the moment the GL call arrives: DirectGLES
+    //    defers and reorders state sync and uploads around draws, and DirectVulkan is usually not
+    //    even recording a command buffer here. A forwarded push/pop would therefore enclose the
+    //    wrong commands, which is a misleading capture rather than a helpful one.
+    //
+    // What the application can rely on is the observable contract: the group stack depth is real
+    // (GL_DEBUG_GROUP_STACK_DEPTH tracks it, and over/underflow raise the errors KHR_debug
+    // specifies), and a label written with glObjectLabel comes back from glGetObjectLabel.
+    void PushDebugGroup(GLenum source, GLuint id, GLsizei length, const GLchar* message);
+    void PopDebugGroup();
+    void DebugMessageInsert(GLenum source, GLenum type, GLuint id, GLenum severity, GLsizei length,
+                            const GLchar* buf);
+    void ObjectLabel(GLenum identifier, GLuint name, GLsizei length, const GLchar* label);
+    void GetObjectLabel(GLenum identifier, GLuint name, GLsizei bufSize, GLsizei* length, GLchar* label);
+
+    // Current depth of the debug group stack, for GL_DEBUG_GROUP_STACK_DEPTH. The base group the
+    // context is created with counts, so this is never below 1 (GL 4.6 core 20.6).
+    GLint GetDebugGroupStackDepth();
+} // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Drawing/GL_Drawing.h

@@ -11,10 +11,31 @@
 
 namespace MobileGL::MG_Impl::GLImpl {
     /* @INSERTION_POINT:FUNCTION_DECLARATION@ */
+    void BeginTransformFeedback(GLenum primitiveMode);
+    void EndTransformFeedback(void);
+    void PauseTransformFeedback(void);
+    void ResumeTransformFeedback(void);
+    void GenTransformFeedbacks(GLsizei n, GLuint* ids);
+    void CreateTransformFeedbacks(GLsizei n, GLuint* ids);
+    void DeleteTransformFeedbacks(GLsizei n, const GLuint* ids);
+    void TransformFeedbackBufferBase(GLuint xfb, GLuint index, GLuint buffer);
+    void TransformFeedbackBufferRange(GLuint xfb, GLuint index, GLuint buffer, GLintptr offset, GLsizeiptr size);
+    void GetTransformFeedbackiv(GLuint xfb, GLenum pname, GLint* param);
+    void GetTransformFeedbacki_v(GLuint xfb, GLenum pname, GLuint index, GLint* param);
+    void GetTransformFeedbacki64_v(GLuint xfb, GLenum pname, GLuint index, GLint64* param);
+    void BindTransformFeedback(GLenum target, GLuint id);
+    GLboolean IsTransformFeedback(GLuint id);
+    void DrawTransformFeedback(GLenum mode, GLuint id);
+    void DrawTransformFeedbackInstanced(GLenum mode, GLuint id, GLsizei instancecount);
+    void DrawTransformFeedbackStream(GLenum mode, GLuint id, GLuint stream);
+    void DrawTransformFeedbackStreamInstanced(GLenum mode, GLuint id, GLuint stream, GLsizei instancecount);
     void DispatchCompute(GLuint numGroupsX, GLuint numGroupsY, GLuint numGroupsZ);
     void DispatchComputeIndirect(GLintptr indirect);
+    void PatchParameteri(GLenum pname, GLint value);
+    void PatchParameterfv(GLenum pname, const GLfloat* values);
     void MemoryBarrier(GLbitfield barriers);
     void MemoryBarrierByRegion(GLbitfield barriers);
+    void TextureBarrier();
     void MultiDrawElementsIndirect(GLenum mode, GLenum type, const void* indirect, GLsizei drawcount, GLsizei stride);
     void MultiDrawArraysIndirect(GLenum mode, const void* indirect, GLsizei drawcount, GLsizei stride);
     void MultiDrawElementsIndirectCount(GLenum mode, GLenum type, const void* indirect, GLintptr drawcount,

MobileGL/MG_Impl/GLImpl/Exporting/Definitions.cpp

@@ -15,21 +15,23 @@
 #include "../Texture/GL_Texture.h"
 #include "../Drawing/GL_Drawing.h"
 #include "../Program/GL_Program.h"
+#include "../Program/GL_ProgramPipeline.h"
 #include "../RenderState/GL_RenderState.h"
 #include "../Framebuffer/GL_Framebuffer.h"
 #include "../VertexArray/GL_VertexArray.h"
 #include "../Sync/GL_Sync.h"
+#include "../Debug/GL_Debug.h"
 #include <MG_State/GLState/Core.h>
 
 #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__) {
@@ -158,7 +160,7 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, ReleaseShaderCompiler) DECLARE_GL_FUNCTION_S
 DECLARE_GL_FUNCTION_HEAD(void, RenderbufferStorage, GLenum target, GLenum internalformat, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, RenderbufferStorage, target, internalformat, width, height)
 DECLARE_GL_FUNCTION_HEAD(void, SampleCoverage, GLfloat value, GLboolean invert) DECLARE_GL_FUNCTION_END_NO_RETURN(void, SampleCoverage, value, invert)
 DECLARE_GL_FUNCTION_HEAD(void, Scissor, GLint x, GLint y, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Scissor, x, y, width, height)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ShaderBinary, GLsizei count, const GLuint* shaders, GLenum binaryformat, const void* binary, GLsizei length) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ShaderBinary, count, shaders, binaryformat, binary, length)
+DECLARE_GL_FUNCTION_HEAD(void, ShaderBinary, GLsizei count, const GLuint* shaders, GLenum binaryformat, const void* binary, GLsizei length) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ShaderBinary, count, shaders, binaryformat, binary, length)
 DECLARE_GL_FUNCTION_HEAD(void, ShaderSource, GLuint shader, GLsizei count, const GLchar* const* string, const GLint* length) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ShaderSource, shader, count, string, length)
 DECLARE_GL_FUNCTION_HEAD(void, StencilFunc, GLenum func, GLint ref, GLuint mask) DECLARE_GL_FUNCTION_END_NO_RETURN(void, StencilFunc, func, ref, mask)
 DECLARE_GL_FUNCTION_HEAD(void, StencilFuncSeparate, GLenum face, GLenum func, GLint ref, GLuint mask) DECLARE_GL_FUNCTION_END_NO_RETURN(void, StencilFuncSeparate, face, func, ref, mask)
@@ -236,12 +238,12 @@ DECLARE_GL_FUNCTION_HEAD(void, DeleteVertexArrays, GLsizei n, const GLuint* arra
 DECLARE_GL_FUNCTION_HEAD(void, GenVertexArrays, GLsizei n, GLuint* arrays) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GenVertexArrays, n, arrays)
 DECLARE_GL_FUNCTION_HEAD(GLboolean, IsVertexArray, GLuint array) DECLARE_GL_FUNCTION_END(GLboolean, IsVertexArray, array)
 DECLARE_GL_FUNCTION_HEAD(void, GetIntegeri_v, GLenum target, GLuint index, GLint* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetIntegeri_v, target, index, data)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, BeginTransformFeedback, GLenum primitiveMode) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BeginTransformFeedback, primitiveMode)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, EndTransformFeedback) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, EndTransformFeedback)
+DECLARE_GL_FUNCTION_HEAD(void, BeginTransformFeedback, GLenum primitiveMode) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BeginTransformFeedback, primitiveMode)
+DECLARE_GL_FUNCTION_HEAD(void, EndTransformFeedback) DECLARE_GL_FUNCTION_END_NO_RETURN(void, EndTransformFeedback)
 DECLARE_GL_FUNCTION_HEAD(void, BindBufferRange, GLenum target, GLuint index, GLuint buffer, GLintptr offset, GLsizeiptr size) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindBufferRange, target, index, buffer, offset, size)
 DECLARE_GL_FUNCTION_HEAD(void, BindBufferBase, GLenum target, GLuint index, GLuint buffer) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindBufferBase, target, index, buffer)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TransformFeedbackVaryings, GLuint program, GLsizei count, const GLchar* const* varyings, GLenum bufferMode) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TransformFeedbackVaryings, program, count, varyings, bufferMode)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetTransformFeedbackVarying, GLuint program, GLuint index, GLsizei bufSize, GLsizei* length, GLsizei* size, GLenum* type, GLchar* name) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetTransformFeedbackVarying, program, index, bufSize, length, size, type, name)
+DECLARE_GL_FUNCTION_HEAD(void, TransformFeedbackVaryings, GLuint program, GLsizei count, const GLchar* const* varyings, GLenum bufferMode) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TransformFeedbackVaryings, program, count, varyings, bufferMode)
+DECLARE_GL_FUNCTION_HEAD(void, GetTransformFeedbackVarying, GLuint program, GLuint index, GLsizei bufSize, GLsizei* length, GLsizei* size, GLenum* type, GLchar* name) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTransformFeedbackVarying, program, index, bufSize, length, size, type, name)
 DECLARE_GL_FUNCTION_HEAD(void, VertexAttribIPointer, GLuint index, GLint size, GLenum type, GLsizei stride, const void* pointer) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexAttribIPointer, index, size, type, stride, pointer)
 DECLARE_GL_FUNCTION_HEAD(void, GetVertexAttribIiv, GLuint index, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetVertexAttribIiv, index, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetVertexAttribIuiv, GLuint index, GLenum pname, GLuint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetVertexAttribIuiv, index, pname, params)
@@ -292,17 +294,17 @@ DECLARE_GL_FUNCTION_HEAD(void, SamplerParameterfv, GLuint sampler, GLenum pname,
 DECLARE_GL_FUNCTION_HEAD(void, GetSamplerParameteriv, GLuint sampler, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetSamplerParameteriv, sampler, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetSamplerParameterfv, GLuint sampler, GLenum pname, GLfloat* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetSamplerParameterfv, sampler, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, VertexAttribDivisor, GLuint index, GLuint divisor) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexAttribDivisor, index, divisor)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, BindTransformFeedback, GLenum target, GLuint id) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BindTransformFeedback, target, id)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, DeleteTransformFeedbacks, GLsizei n, const GLuint* ids) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DeleteTransformFeedbacks, n, ids)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GenTransformFeedbacks, GLsizei n, GLuint* ids) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GenTransformFeedbacks, n, ids)
-DECLARE_GL_FUNCTION_STUB_HEAD(GLboolean, IsTransformFeedback, GLuint id) DECLARE_GL_FUNCTION_STUB_END(GLboolean, IsTransformFeedback, id)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, PauseTransformFeedback) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PauseTransformFeedback)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ResumeTransformFeedback) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ResumeTransformFeedback)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetProgramBinary, GLuint program, GLsizei bufSize, GLsizei* length, GLenum* binaryFormat, void* binary) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetProgramBinary, program, bufSize, length, binaryFormat, binary)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramBinary, GLuint program, GLenum binaryFormat, const void* binary, GLsizei length) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramBinary, program, binaryFormat, binary, length)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramParameteri, GLuint program, GLenum pname, GLint value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramParameteri, program, pname, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, InvalidateFramebuffer, GLenum target, GLsizei numAttachments, const GLenum* attachments) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, InvalidateFramebuffer, target, numAttachments, attachments)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, InvalidateSubFramebuffer, GLenum target, GLsizei numAttachments, const GLenum* attachments, GLint x, GLint y, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, InvalidateSubFramebuffer, target, numAttachments, attachments, x, y, width, height)
+DECLARE_GL_FUNCTION_HEAD(void, BindTransformFeedback, GLenum target, GLuint id) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindTransformFeedback, target, id)
+DECLARE_GL_FUNCTION_HEAD(void, DeleteTransformFeedbacks, GLsizei n, const GLuint* ids) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DeleteTransformFeedbacks, n, ids)
+DECLARE_GL_FUNCTION_HEAD(void, GenTransformFeedbacks, GLsizei n, GLuint* ids) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GenTransformFeedbacks, n, ids)
+DECLARE_GL_FUNCTION_HEAD(GLboolean, IsTransformFeedback, GLuint id) DECLARE_GL_FUNCTION_END(GLboolean, IsTransformFeedback, id)
+DECLARE_GL_FUNCTION_HEAD(void, PauseTransformFeedback) DECLARE_GL_FUNCTION_END_NO_RETURN(void, PauseTransformFeedback)
+DECLARE_GL_FUNCTION_HEAD(void, ResumeTransformFeedback) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ResumeTransformFeedback)
+DECLARE_GL_FUNCTION_HEAD(void, GetProgramBinary, GLuint program, GLsizei bufSize, GLsizei* length, GLenum* binaryFormat, void* binary) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetProgramBinary, program, bufSize, length, binaryFormat, binary)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramBinary, GLuint program, GLenum binaryFormat, const void* binary, GLsizei length) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramBinary, program, binaryFormat, binary, length)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramParameteri, GLuint program, GLenum pname, GLint value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramParameteri, program, pname, value)
+DECLARE_GL_FUNCTION_HEAD(void, InvalidateFramebuffer, GLenum target, GLsizei numAttachments, const GLenum* attachments) DECLARE_GL_FUNCTION_END_NO_RETURN(void, InvalidateFramebuffer, target, numAttachments, attachments)
+DECLARE_GL_FUNCTION_HEAD(void, InvalidateSubFramebuffer, GLenum target, GLsizei numAttachments, const GLenum* attachments, GLint x, GLint y, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, InvalidateSubFramebuffer, target, numAttachments, attachments, x, y, width, height)
 DECLARE_GL_FUNCTION_HEAD(void, TexStorage2D, GLenum target, GLsizei levels, GLenum internalformat, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TexStorage2D, target, levels, internalformat, width, height)
 DECLARE_GL_FUNCTION_HEAD(void, TexStorage3D, GLenum target, GLsizei levels, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TexStorage3D, target, levels, internalformat, width, height, depth)
 DECLARE_GL_FUNCTION_HEAD(void, GetInternalformativ, GLenum target, GLenum internalformat, GLenum pname, GLsizei bufSize, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetInternalformativ, target, internalformat, pname, bufSize, params)
@@ -310,21 +312,21 @@ DECLARE_GL_FUNCTION_HEAD(void, DispatchCompute, GLuint num_groups_x, GLuint num_
 DECLARE_GL_FUNCTION_HEAD(void, DispatchComputeIndirect, GLintptr indirect) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DispatchComputeIndirect, indirect)
 DECLARE_GL_FUNCTION_HEAD(void, DrawArraysIndirect, GLenum mode, const void* indirect) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawArraysIndirect, mode, indirect)
 DECLARE_GL_FUNCTION_HEAD(void, DrawElementsIndirect, GLenum mode, GLenum type, const void* indirect) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawElementsIndirect, mode, type, indirect)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, FramebufferParameteri, GLenum target, GLenum pname, GLint param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, FramebufferParameteri, target, pname, param)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetFramebufferParameteriv, GLenum target, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetFramebufferParameteriv, target, pname, params)
+DECLARE_GL_FUNCTION_HEAD(void, FramebufferParameteri, GLenum target, GLenum pname, GLint param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, FramebufferParameteri, target, pname, param)
+DECLARE_GL_FUNCTION_HEAD(void, GetFramebufferParameteriv, GLenum target, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetFramebufferParameteriv, target, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetProgramInterfaceiv, GLuint program, GLenum programInterface, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetProgramInterfaceiv, program, programInterface, pname, params)
 DECLARE_GL_FUNCTION_HEAD(GLuint, GetProgramResourceIndex, GLuint program, GLenum programInterface, const GLchar* name) DECLARE_GL_FUNCTION_END(GLuint, GetProgramResourceIndex, program, programInterface, name)
 DECLARE_GL_FUNCTION_HEAD(void, GetProgramResourceName, GLuint program, GLenum programInterface, GLuint index, GLsizei bufSize, GLsizei* length, GLchar* name) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetProgramResourceName, program, programInterface, index, bufSize, length, name)
 DECLARE_GL_FUNCTION_HEAD(void, GetProgramResourceiv, GLuint program, GLenum programInterface, GLuint index, GLsizei propCount, const GLenum* props, GLsizei bufSize, GLsizei* length, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetProgramResourceiv, program, programInterface, index, propCount, props, bufSize, length, params)
 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_STUB_HEAD(void, UseProgramStages, GLuint pipeline, GLbitfield stages, GLuint program) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UseProgramStages, pipeline, stages, program)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ActiveShaderProgram, GLuint pipeline, GLuint program) DECLARE_GL_FUNCTION_STUB_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_STUB_HEAD(void, BindProgramPipeline, GLuint pipeline) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BindProgramPipeline, pipeline)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, DeleteProgramPipelines, GLsizei n, const GLuint* pipelines) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DeleteProgramPipelines, n, pipelines)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GenProgramPipelines, GLsizei n, GLuint* pipelines) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GenProgramPipelines, n, pipelines)
-DECLARE_GL_FUNCTION_STUB_HEAD(GLboolean, IsProgramPipeline, GLuint pipeline) DECLARE_GL_FUNCTION_STUB_END(GLboolean, IsProgramPipeline, pipeline)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetProgramPipelineiv, GLuint pipeline, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetProgramPipelineiv, pipeline, pname, params)
+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_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)
+DECLARE_GL_FUNCTION_HEAD(GLboolean, IsProgramPipeline, GLuint pipeline) DECLARE_GL_FUNCTION_END(GLboolean, IsProgramPipeline, pipeline)
+DECLARE_GL_FUNCTION_HEAD(void, GetProgramPipelineiv, GLuint pipeline, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetProgramPipelineiv, pipeline, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform1i, GLuint program, GLint location, GLint v0) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform1i, program, location, v0)
 DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform2i, GLuint program, GLint location, GLint v0, GLint v1) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform2i, program, location, v0, v1)
 DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform3i, GLuint program, GLint location, GLint v0, GLint v1, GLint v2) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform3i, program, location, v0, v1, v2)
@@ -358,8 +360,8 @@ DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix2x4fv, GLuint program, GLint
 DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix4x2fv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLfloat* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix4x2fv, program, location, count, transpose, value)
 DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix3x4fv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLfloat* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix3x4fv, program, location, count, transpose, value)
 DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix4x3fv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLfloat* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix4x3fv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ValidateProgramPipeline, GLuint pipeline) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ValidateProgramPipeline, pipeline)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetProgramPipelineInfoLog, GLuint pipeline, GLsizei bufSize, GLsizei* length, GLchar* infoLog) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetProgramPipelineInfoLog, pipeline, bufSize, length, infoLog)
+DECLARE_GL_FUNCTION_HEAD(void, ValidateProgramPipeline, GLuint pipeline) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ValidateProgramPipeline, pipeline)
+DECLARE_GL_FUNCTION_HEAD(void, GetProgramPipelineInfoLog, GLuint pipeline, GLsizei bufSize, GLsizei* length, GLchar* infoLog) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetProgramPipelineInfoLog, pipeline, bufSize, length, infoLog)
 DECLARE_GL_FUNCTION_HEAD(void, BindImageTexture, GLuint unit, GLuint texture, GLint level, GLboolean layered, GLint layer, GLenum access, GLenum format) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindImageTexture, unit, texture, level, layered, layer, access, format)
 DECLARE_GL_FUNCTION_HEAD(void, GetBooleani_v, GLenum target, GLuint index, GLboolean* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetBooleani_v, target, index, data)
 DECLARE_GL_FUNCTION_HEAD(void, MemoryBarrier, GLbitfield barriers) DECLARE_GL_FUNCTION_END_NO_RETURN(void, MemoryBarrier, barriers)
@@ -377,27 +379,13 @@ DECLARE_GL_FUNCTION_HEAD(void, VertexBindingDivisor, GLuint bindingindex, GLuint
 DECLARE_GL_FUNCTION_STUB_HEAD(void, BlendBarrier) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BlendBarrier)
 DECLARE_GL_FUNCTION_HEAD(void, CopyImageSubData, GLuint srcName, GLenum srcTarget, GLint srcLevel, GLint srcX, GLint srcY, GLint srcZ, GLuint dstName, GLenum dstTarget, GLint dstLevel, GLint dstX, GLint dstY, GLint dstZ, GLsizei srcWidth, GLsizei srcHeight, GLsizei srcDepth) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CopyImageSubData, srcName, srcTarget, srcLevel, srcX, srcY, srcZ, dstName, dstTarget, dstLevel, dstX, dstY, dstZ, srcWidth, srcHeight, srcDepth)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, DebugMessageControl, GLenum source, GLenum type, GLenum severity, GLsizei count, const GLuint* ids, GLboolean enabled) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DebugMessageControl, source, type, severity, count, ids, enabled)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, DebugMessageInsert, GLenum source, GLenum type, GLuint id, GLenum severity, GLsizei length, const GLchar* buf) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DebugMessageInsert, source, type, id, severity, length, buf)
+DECLARE_GL_FUNCTION_HEAD(void, DebugMessageInsert, GLenum source, GLenum type, GLuint id, GLenum severity, GLsizei length, const GLchar* buf) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DebugMessageInsert, source, type, id, severity, length, buf)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, DebugMessageCallback, GLDEBUGPROC callback, const void* userParam) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DebugMessageCallback, callback, userParam)
 DECLARE_GL_FUNCTION_STUB_HEAD(GLuint, GetDebugMessageLog, GLuint count, GLsizei bufSize, GLenum* sources, GLenum* types, GLuint* ids, GLenum* severities, GLsizei* lengths, GLchar* messageLog) DECLARE_GL_FUNCTION_STUB_END(GLuint, GetDebugMessageLog, count, bufSize, sources, types, ids, severities, lengths, messageLog)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, PushDebugGroup, GLenum source, GLuint id, GLsizei length, const GLchar* message) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PushDebugGroup, source, id, length, message)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, PopDebugGroup) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PopDebugGroup)
-MOBILEGL_GL_API void glObjectLabel(GLenum identifier, GLuint name, GLsizei length, const GLchar* label) {
-    (void)identifier;
-    (void)name;
-    (void)length;
-    (void)label;
-}
-MOBILEGL_GL_API void glGetObjectLabel(GLenum identifier, GLuint name, GLsizei bufSize, GLsizei* length, GLchar* label) {
-    (void)identifier;
-    (void)name;
-    if (length) {
-        *length = 0;
-    }
-    if (label && bufSize > 0) {
-        label[0] = '\0';
-    }
-}
+DECLARE_GL_FUNCTION_HEAD(void, PushDebugGroup, GLenum source, GLuint id, GLsizei length, const GLchar* message) DECLARE_GL_FUNCTION_END_NO_RETURN(void, PushDebugGroup, source, id, length, message)
+DECLARE_GL_FUNCTION_HEAD(void, PopDebugGroup) DECLARE_GL_FUNCTION_END_NO_RETURN(void, PopDebugGroup)
+DECLARE_GL_FUNCTION_HEAD(void, ObjectLabel, GLenum identifier, GLuint name, GLsizei length, const GLchar* label) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ObjectLabel, identifier, name, length, label)
+DECLARE_GL_FUNCTION_HEAD(void, GetObjectLabel, GLenum identifier, GLuint name, GLsizei bufSize, GLsizei* length, GLchar* label) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetObjectLabel, identifier, name, bufSize, length, label)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, ObjectPtrLabel, const void* ptr, GLsizei length, const GLchar* label) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ObjectPtrLabel, ptr, length, label)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetObjectPtrLabel, const void* ptr, GLsizei bufSize, GLsizei* length, GLchar* label) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetObjectPtrLabel, ptr, bufSize, length, label)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetPointerv, GLenum pname, void** params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetPointerv, pname, params)
@@ -418,13 +406,13 @@ DECLARE_GL_FUNCTION_HEAD(void, DrawRangeElementsBaseVertex, GLenum mode, GLuint
 DECLARE_GL_FUNCTION_HEAD(void, DrawElementsInstancedBaseVertex, GLenum mode, GLsizei count, GLenum type, const void* indices, GLsizei instancecount, GLint basevertex) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawElementsInstancedBaseVertex, mode, count, type, indices, instancecount, basevertex)
 DECLARE_GL_FUNCTION_HEAD(void, FramebufferTexture, GLenum target, GLenum attachment, GLuint texture, GLint level) DECLARE_GL_FUNCTION_END_NO_RETURN(void, FramebufferTexture, target, attachment, texture, level)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PrimitiveBoundingBox, GLfloat minX, GLfloat minY, GLfloat minZ, GLfloat minW, GLfloat maxX, GLfloat maxY, GLfloat maxZ, GLfloat maxW) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PrimitiveBoundingBox, minX, minY, minZ, minW, maxX, maxY, maxZ, maxW)
-DECLARE_GL_FUNCTION_STUB_HEAD(GLenum, GetGraphicsResetStatus) DECLARE_GL_FUNCTION_STUB_END(GLenum, GetGraphicsResetStatus)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ReadnPixels, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, GLsizei bufSize, void* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ReadnPixels, x, y, width, height, format, type, bufSize, data)
+DECLARE_GL_FUNCTION_HEAD(GLenum, GetGraphicsResetStatus) DECLARE_GL_FUNCTION_END(GLenum, GetGraphicsResetStatus)
+DECLARE_GL_FUNCTION_HEAD(void, ReadnPixels, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, GLsizei bufSize, void* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ReadnPixels, x, y, width, height, format, type, bufSize, data)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnUniformfv, GLuint program, GLint location, GLsizei bufSize, GLfloat* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetnUniformfv, program, location, bufSize, params)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnUniformiv, GLuint program, GLint location, GLsizei bufSize, GLint* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetnUniformiv, program, location, bufSize, params)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnUniformuiv, GLuint program, GLint location, GLsizei bufSize, GLuint* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetnUniformuiv, program, location, bufSize, params)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, MinSampleShading, GLfloat value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, MinSampleShading, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, PatchParameteri, GLenum pname, GLint value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PatchParameteri, pname, value)
+DECLARE_GL_FUNCTION_HEAD(void, MinSampleShading, GLfloat value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, MinSampleShading, value)
+DECLARE_GL_FUNCTION_HEAD(void, PatchParameteri, GLenum pname, GLint value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, PatchParameteri, pname, value)
 DECLARE_GL_FUNCTION_HEAD(void, TexParameterIiv, GLenum target, GLenum pname, const GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TexParameterIiv, target, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, TexParameterIuiv, GLenum target, GLenum pname, const GLuint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TexParameterIuiv, target, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetTexParameterIiv, GLenum target, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTexParameterIiv, target, pname, params)
@@ -434,7 +422,7 @@ DECLARE_GL_FUNCTION_HEAD(void, SamplerParameterIuiv, GLuint sampler, GLenum pnam
 DECLARE_GL_FUNCTION_HEAD(void, GetSamplerParameterIiv, GLuint sampler, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetSamplerParameterIiv, sampler, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetSamplerParameterIuiv, GLuint sampler, GLenum pname, GLuint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetSamplerParameterIuiv, sampler, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, TexBuffer, GLenum target, GLenum internalformat, GLuint buffer) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TexBuffer, target, internalformat, buffer)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TexBufferRange, GLenum target, GLenum internalformat, GLuint buffer, GLintptr offset, GLsizeiptr size) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TexBufferRange, target, internalformat, buffer, offset, size)
+DECLARE_GL_FUNCTION_HEAD(void, TexBufferRange, GLenum target, GLenum internalformat, GLuint buffer, GLintptr offset, GLsizeiptr size) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TexBufferRange, target, internalformat, buffer, offset, size)
 DECLARE_GL_FUNCTION_HEAD(void, TexStorage3DMultisample, GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLboolean fixedsamplelocations) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TexStorage3DMultisample, target, samples, internalformat, width, height, depth, fixedsamplelocations)
 DECLARE_GL_FUNCTION_HEAD(void*, MapBufferRange, GLenum target, GLintptr offset, GLsizeiptr length, GLbitfield access) DECLARE_GL_FUNCTION_END(void*, MapBufferRange, target, offset, length, access)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, ClearIndex, GLfloat c) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ClearIndex, c)
@@ -724,8 +712,8 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, LoadName, GLuint name) DECLARE_GL_FUNCTION_S
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PushName, GLuint name) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PushName, name)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PopName) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PopName)
 DECLARE_GL_FUNCTION_HEAD(void, ClampColor, GLenum target, GLenum clamp) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClampColor, target, clamp)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, BeginConditionalRender, GLuint id, GLenum mode) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BeginConditionalRender, id, mode)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, EndConditionalRender, void) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, EndConditionalRender)
+DECLARE_GL_FUNCTION_HEAD(void, BeginConditionalRender, GLuint id, GLenum mode) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BeginConditionalRender, id, mode)
+DECLARE_GL_FUNCTION_HEAD(void, EndConditionalRender) DECLARE_GL_FUNCTION_END_NO_RETURN(void, EndConditionalRender)
 DECLARE_GL_FUNCTION_HEAD(void, VertexAttribI1i, GLuint index, GLint x) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexAttribI1i, index, x)
 DECLARE_GL_FUNCTION_HEAD(void, VertexAttribI2i, GLuint index, GLint x, GLint y) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexAttribI2i, index, x, y)
 DECLARE_GL_FUNCTION_HEAD(void, VertexAttribI3i, GLuint index, GLint x, GLint y, GLint z) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexAttribI3i, index, x, y, z)
@@ -909,24 +897,24 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, ColorP4ui, GLenum type, GLuint color) DECLAR
 DECLARE_GL_FUNCTION_STUB_HEAD(void, ColorP4uiv, GLenum type, const GLuint* color) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ColorP4uiv, type, color)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, SecondaryColorP3ui, GLenum type, GLuint color) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, SecondaryColorP3ui, type, color)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, SecondaryColorP3uiv, GLenum type, const GLuint* color) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, SecondaryColorP3uiv, type, color)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform1d, GLint location, GLdouble x) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform1d, location, x)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform2d, GLint location, GLdouble x, GLdouble y) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform2d, location, x, y)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform3d, GLint location, GLdouble x, GLdouble y, GLdouble z) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform3d, location, x, y, z)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform4d, GLint location, GLdouble x, GLdouble y, GLdouble z, GLdouble w) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform4d, location, x, y, z, w)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform1dv, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform1dv, location, count, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform2dv, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform2dv, location, count, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform3dv, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform3dv, location, count, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform4dv, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform4dv, location, count, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix2dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix2dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix3dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix3dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix4dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix4dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix2x3dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix2x3dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix2x4dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix2x4dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix3x2dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix3x2dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix3x4dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix3x4dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix4x2dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix4x2dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformMatrix4x3dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformMatrix4x3dv, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetUniformdv, GLuint program, GLint location, GLdouble* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetUniformdv, program, location, params)
+DECLARE_GL_FUNCTION_HEAD(void, Uniform1d, GLint location, GLdouble x) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Uniform1d, location, x)
+DECLARE_GL_FUNCTION_HEAD(void, Uniform2d, GLint location, GLdouble x, GLdouble y) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Uniform2d, location, x, y)
+DECLARE_GL_FUNCTION_HEAD(void, Uniform3d, GLint location, GLdouble x, GLdouble y, GLdouble z) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Uniform3d, location, x, y, z)
+DECLARE_GL_FUNCTION_HEAD(void, Uniform4d, GLint location, GLdouble x, GLdouble y, GLdouble z, GLdouble w) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Uniform4d, location, x, y, z, w)
+DECLARE_GL_FUNCTION_HEAD(void, Uniform1dv, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Uniform1dv, location, count, value)
+DECLARE_GL_FUNCTION_HEAD(void, Uniform2dv, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Uniform2dv, location, count, value)
+DECLARE_GL_FUNCTION_HEAD(void, Uniform3dv, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Uniform3dv, location, count, value)
+DECLARE_GL_FUNCTION_HEAD(void, Uniform4dv, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, Uniform4dv, location, count, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix2dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix2dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix3dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix3dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix4dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix4dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix2x3dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix2x3dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix2x4dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix2x4dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix3x2dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix3x2dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix3x4dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix3x4dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix4x2dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix4x2dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, UniformMatrix4x3dv, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, UniformMatrix4x3dv, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, GetUniformdv, GLuint program, GLint location, GLdouble* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetUniformdv, program, location, params)
 DECLARE_GL_FUNCTION_STUB_HEAD(GLint, GetSubroutineUniformLocation, GLuint program, GLenum shadertype, const GLchar* name) DECLARE_GL_FUNCTION_STUB_END(GLint, GetSubroutineUniformLocation, program, shadertype, name)
 DECLARE_GL_FUNCTION_STUB_HEAD(GLuint, GetSubroutineIndex, GLuint program, GLenum shadertype, const GLchar* name) DECLARE_GL_FUNCTION_STUB_END(GLuint, GetSubroutineIndex, program, shadertype, name)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetActiveSubroutineUniformiv, GLuint program, GLenum shadertype, GLuint index, GLenum pname, GLint* values) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetActiveSubroutineUniformiv, program, shadertype, index, pname, values)
@@ -935,29 +923,29 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, GetActiveSubroutineName, GLuint program, GLe
 DECLARE_GL_FUNCTION_STUB_HEAD(void, UniformSubroutinesuiv, GLenum shadertype, GLsizei count, const GLuint* indices) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, UniformSubroutinesuiv, shadertype, count, indices)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetUniformSubroutineuiv, GLenum shadertype, GLint location, GLuint* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetUniformSubroutineuiv, shadertype, location, params)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetProgramStageiv, GLuint program, GLenum shadertype, GLenum pname, GLint* values) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetProgramStageiv, program, shadertype, pname, values)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, PatchParameterfv, GLenum pname, const GLfloat* values) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PatchParameterfv, pname, values)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, DrawTransformFeedback, GLenum mode, GLuint id) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DrawTransformFeedback, mode, id)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, DrawTransformFeedbackStream, GLenum mode, GLuint id, GLuint stream) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DrawTransformFeedbackStream, mode, id, stream)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, BeginQueryIndexed, GLenum target, GLuint index, GLuint id) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BeginQueryIndexed, target, index, id)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, EndQueryIndexed, GLenum target, GLuint index) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, EndQueryIndexed, target, index)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetQueryIndexediv, GLenum target, GLuint index, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetQueryIndexediv, target, index, pname, params)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniform1d, GLuint program, GLint location, GLdouble v0) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniform1d, program, location, v0)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniform1dv, GLuint program, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniform1dv, program, location, count, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniform2d, GLuint program, GLint location, GLdouble v0, GLdouble v1) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniform2d, program, location, v0, v1)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniform2dv, GLuint program, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniform2dv, program, location, count, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniform3d, GLuint program, GLint location, GLdouble v0, GLdouble v1, GLdouble v2) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniform3d, program, location, v0, v1, v2)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniform3dv, GLuint program, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniform3dv, program, location, count, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniform4d, GLuint program, GLint location, GLdouble v0, GLdouble v1, GLdouble v2, GLdouble v3) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniform4d, program, location, v0, v1, v2, v3)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniform4dv, GLuint program, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniform4dv, program, location, count, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix2dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix2dv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix3dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix3dv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix4dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix4dv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix2x3dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix2x3dv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix3x2dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix3x2dv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix2x4dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix2x4dv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix4x2dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix4x2dv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix3x4dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix3x4dv, program, location, count, transpose, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ProgramUniformMatrix4x3dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ProgramUniformMatrix4x3dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, PatchParameterfv, GLenum pname, const GLfloat* values) DECLARE_GL_FUNCTION_END_NO_RETURN(void, PatchParameterfv, pname, values)
+DECLARE_GL_FUNCTION_HEAD(void, DrawTransformFeedback, GLenum mode, GLuint id) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawTransformFeedback, mode, id)
+DECLARE_GL_FUNCTION_HEAD(void, DrawTransformFeedbackStream, GLenum mode, GLuint id, GLuint stream) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DrawTransformFeedbackStream, mode, id, stream)
+DECLARE_GL_FUNCTION_HEAD(void, BeginQueryIndexed, GLenum target, GLuint index, GLuint id) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BeginQueryIndexed, target, index, id)
+DECLARE_GL_FUNCTION_HEAD(void, EndQueryIndexed, GLenum target, GLuint index) DECLARE_GL_FUNCTION_END_NO_RETURN(void, EndQueryIndexed, target, index)
+DECLARE_GL_FUNCTION_HEAD(void, GetQueryIndexediv, GLenum target, GLuint index, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetQueryIndexediv, target, index, pname, params)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform1d, GLuint program, GLint location, GLdouble v0) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform1d, program, location, v0)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform1dv, GLuint program, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform1dv, program, location, count, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform2d, GLuint program, GLint location, GLdouble v0, GLdouble v1) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform2d, program, location, v0, v1)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform2dv, GLuint program, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform2dv, program, location, count, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform3d, GLuint program, GLint location, GLdouble v0, GLdouble v1, GLdouble v2) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform3d, program, location, v0, v1, v2)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform3dv, GLuint program, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform3dv, program, location, count, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform4d, GLuint program, GLint location, GLdouble v0, GLdouble v1, GLdouble v2, GLdouble v3) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform4d, program, location, v0, v1, v2, v3)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniform4dv, GLuint program, GLint location, GLsizei count, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniform4dv, program, location, count, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix2dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix2dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix3dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix3dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix4dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix4dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix2x3dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix2x3dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix3x2dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix3x2dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix2x4dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix2x4dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix4x2dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix4x2dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix3x4dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix3x4dv, program, location, count, transpose, value)
+DECLARE_GL_FUNCTION_HEAD(void, ProgramUniformMatrix4x3dv, GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProgramUniformMatrix4x3dv, program, location, count, transpose, value)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, VertexAttribL1d, GLuint index, GLdouble x) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, VertexAttribL1d, index, x)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, VertexAttribL2d, GLuint index, GLdouble x, GLdouble y) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, VertexAttribL2d, index, x, y)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, VertexAttribL3d, GLuint index, GLdouble x, GLdouble y, GLdouble z) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, VertexAttribL3d, index, x, y, z)
@@ -968,24 +956,24 @@ 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_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_STUB_HEAD(void, ViewportIndexedfv, GLuint index, const GLfloat* v) DECLARE_GL_FUNCTION_STUB_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_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_STUB_HEAD(void, ScissorIndexedv, GLuint index, const GLint* v) DECLARE_GL_FUNCTION_STUB_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_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_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_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, ViewportArrayv, GLuint first, GLsizei count, const GLfloat* v) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ViewportArrayv, first, count, v)
+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_HEAD(void, ViewportIndexedfv, GLuint index, const GLfloat* v) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ViewportIndexedfv, index, 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_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_HEAD(void, ScissorIndexedv, GLuint index, const GLint* v) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ScissorIndexedv, index, 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_HEAD(void, DepthRangeIndexed, GLuint index, GLdouble n, GLdouble f) DECLARE_GL_FUNCTION_END_NO_RETURN(void, DepthRangeIndexed, index, n, f)
+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_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_STUB_HEAD(void, DrawTransformFeedbackInstanced, GLenum mode, GLuint id, GLsizei instancecount) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DrawTransformFeedbackInstanced, mode, id, instancecount)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, DrawTransformFeedbackStreamInstanced, GLenum mode, GLuint id, GLuint stream, GLsizei instancecount) DECLARE_GL_FUNCTION_STUB_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_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, GetActiveAtomicCounterBufferiv, GLuint program, GLuint bufferIndex, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_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_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_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)
@@ -995,24 +983,24 @@ DECLARE_GL_FUNCTION_HEAD(void, MultiDrawArraysIndirect, GLenum mode, const void*
 DECLARE_GL_FUNCTION_HEAD(void, MultiDrawElementsIndirect, GLenum mode, GLenum type, const void* indirect, GLsizei drawcount, GLsizei stride) DECLARE_GL_FUNCTION_END_NO_RETURN(void, MultiDrawElementsIndirect, mode, type, indirect, drawcount, stride)
 DECLARE_GL_FUNCTION_HEAD(GLint, GetProgramResourceLocationIndex, GLuint program, GLenum programInterface, const GLchar* name) DECLARE_GL_FUNCTION_END(GLint, GetProgramResourceLocationIndex, program, programInterface, name)
 DECLARE_GL_FUNCTION_HEAD(void, ShaderStorageBlockBinding, GLuint program, GLuint storageBlockIndex, GLuint storageBlockBinding) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ShaderStorageBlockBinding, program, storageBlockIndex, storageBlockBinding)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TextureView, GLuint texture, GLenum target, GLuint origtexture, GLenum internalformat, GLuint minlevel, GLuint numlevels, GLuint minlayer, GLuint numlayers) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TextureView, texture, target, origtexture, internalformat, minlevel, numlevels, minlayer, numlayers)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, VertexAttribLFormat, GLuint attribindex, GLint size, GLenum type, GLuint relativeoffset) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, VertexAttribLFormat, attribindex, size, type, relativeoffset)
+DECLARE_GL_FUNCTION_HEAD(void, TextureView, GLuint texture, GLenum target, GLuint origtexture, GLenum internalformat, GLuint minlevel, GLuint numlevels, GLuint minlayer, GLuint numlayers) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureView, texture, target, origtexture, internalformat, minlevel, numlevels, minlayer, numlayers)
+DECLARE_GL_FUNCTION_HEAD(void, VertexAttribLFormat, GLuint attribindex, GLint size, GLenum type, GLuint relativeoffset) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexAttribLFormat, attribindex, size, type, relativeoffset)
 DECLARE_GL_FUNCTION_HEAD(void, BufferStorage, GLenum target, GLsizeiptr size, const void* data, GLbitfield flags) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BufferStorage, target, size, data, flags)
 DECLARE_GL_FUNCTION_HEAD(void, ClearTexImage, GLuint texture, GLint level, GLenum format, GLenum type, const void* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClearTexImage, texture, level, format, type, data)
 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_STUB_HEAD(void, CreateTransformFeedbacks, GLsizei n, GLuint* ids) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CreateTransformFeedbacks, n, ids)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TransformFeedbackBufferBase, GLuint xfb, GLuint index, GLuint buffer) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TransformFeedbackBufferBase, xfb, index, buffer)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TransformFeedbackBufferRange, GLuint xfb, GLuint index, GLuint buffer, GLintptr offset, GLsizeiptr size) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TransformFeedbackBufferRange, xfb, index, buffer, offset, size)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetTransformFeedbackiv, GLuint xfb, GLenum pname, GLint* param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetTransformFeedbackiv, xfb, pname, param)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetTransformFeedbacki_v, GLuint xfb, GLenum pname, GLuint index, GLint* param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetTransformFeedbacki_v, xfb, pname, index, param)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetTransformFeedbacki64_v, GLuint xfb, GLenum pname, GLuint index, GLint64* param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetTransformFeedbacki64_v, xfb, pname, index, param)
+DECLARE_GL_FUNCTION_HEAD(void, ClipControl, GLenum origin, GLenum depth) DECLARE_GL_FUNCTION_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)
+DECLARE_GL_FUNCTION_HEAD(void, TransformFeedbackBufferBase, GLuint xfb, GLuint index, GLuint buffer) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TransformFeedbackBufferBase, xfb, index, buffer)
+DECLARE_GL_FUNCTION_HEAD(void, TransformFeedbackBufferRange, GLuint xfb, GLuint index, GLuint buffer, GLintptr offset, GLsizeiptr size) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TransformFeedbackBufferRange, xfb, index, buffer, offset, size)
+DECLARE_GL_FUNCTION_HEAD(void, GetTransformFeedbackiv, GLuint xfb, GLenum pname, GLint* param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTransformFeedbackiv, xfb, pname, param)
+DECLARE_GL_FUNCTION_HEAD(void, GetTransformFeedbacki_v, GLuint xfb, GLenum pname, GLuint index, GLint* param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTransformFeedbacki_v, xfb, pname, index, param)
+DECLARE_GL_FUNCTION_HEAD(void, GetTransformFeedbacki64_v, GLuint xfb, GLenum pname, GLuint index, GLint64* param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTransformFeedbacki64_v, xfb, pname, index, param)
 DECLARE_GL_FUNCTION_HEAD(void, CreateBuffers, GLsizei n, GLuint* buffers) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CreateBuffers, n, buffers)
 DECLARE_GL_FUNCTION_HEAD(void, NamedBufferStorage, GLuint buffer, GLsizeiptr size, const void* data, GLbitfield flags) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedBufferStorage, buffer, size, data, flags)
 DECLARE_GL_FUNCTION_HEAD(void, NamedBufferData, GLuint buffer, GLsizeiptr size, const void* data, GLenum usage) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedBufferData, buffer, size, data, usage)
@@ -1025,32 +1013,32 @@ DECLARE_GL_FUNCTION_HEAD(void, FlushMappedNamedBufferRange, GLuint buffer, GLint
 DECLARE_GL_FUNCTION_HEAD(void, GetNamedBufferParameteriv, GLuint buffer, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetNamedBufferParameteriv, buffer, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetNamedBufferParameteri64v, GLuint buffer, GLenum pname, GLint64* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetNamedBufferParameteri64v, buffer, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetNamedBufferPointerv, GLuint buffer, GLenum pname, void** params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetNamedBufferPointerv, buffer, pname, params)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetNamedBufferSubData, GLuint buffer, GLintptr offset, GLsizeiptr size, void* data) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetNamedBufferSubData, buffer, offset, size, data)
+DECLARE_GL_FUNCTION_HEAD(void, GetNamedBufferSubData, GLuint buffer, GLintptr offset, GLsizeiptr size, void* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetNamedBufferSubData, buffer, offset, size, data)
 DECLARE_GL_FUNCTION_HEAD(void, CreateFramebuffers, GLsizei n, GLuint* framebuffers) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CreateFramebuffers, n, framebuffers)
 DECLARE_GL_FUNCTION_HEAD(void, NamedFramebufferRenderbuffer, GLuint framebuffer, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedFramebufferRenderbuffer, framebuffer, attachment, renderbuffertarget, renderbuffer)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, NamedFramebufferParameteri, GLuint framebuffer, GLenum pname, GLint param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, NamedFramebufferParameteri, framebuffer, pname, param)
+DECLARE_GL_FUNCTION_HEAD(void, NamedFramebufferParameteri, GLuint framebuffer, GLenum pname, GLint param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedFramebufferParameteri, framebuffer, pname, param)
 DECLARE_GL_FUNCTION_HEAD(void, NamedFramebufferTexture, GLuint framebuffer, GLenum attachment, GLuint texture, GLint level) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedFramebufferTexture, framebuffer, attachment, texture, level)
 DECLARE_GL_FUNCTION_HEAD(void, NamedFramebufferTextureLayer, GLuint framebuffer, GLenum attachment, GLuint texture, GLint level, GLint layer) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedFramebufferTextureLayer, framebuffer, attachment, texture, level, layer)
 DECLARE_GL_FUNCTION_HEAD(void, NamedFramebufferDrawBuffer, GLuint framebuffer, GLenum buf) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedFramebufferDrawBuffer, framebuffer, buf)
 DECLARE_GL_FUNCTION_HEAD(void, NamedFramebufferDrawBuffers, GLuint framebuffer, GLsizei n, const GLenum* bufs) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedFramebufferDrawBuffers, framebuffer, n, bufs)
 DECLARE_GL_FUNCTION_HEAD(void, NamedFramebufferReadBuffer, GLuint framebuffer, GLenum src) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedFramebufferReadBuffer, framebuffer, src)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, InvalidateNamedFramebufferData, GLuint framebuffer, GLsizei numAttachments, const GLenum* attachments) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, InvalidateNamedFramebufferData, framebuffer, numAttachments, attachments)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, InvalidateNamedFramebufferSubData, GLuint framebuffer, GLsizei numAttachments, const GLenum* attachments, GLint x, GLint y, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, InvalidateNamedFramebufferSubData, framebuffer, numAttachments, attachments, x, y, width, height)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ClearNamedFramebufferiv, GLuint framebuffer, GLenum buffer, GLint drawbuffer, const GLint* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ClearNamedFramebufferiv, framebuffer, buffer, drawbuffer, value)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, ClearNamedFramebufferuiv, GLuint framebuffer, GLenum buffer, GLint drawbuffer, const GLuint* value) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ClearNamedFramebufferuiv, framebuffer, buffer, drawbuffer, value)
+DECLARE_GL_FUNCTION_HEAD(void, InvalidateNamedFramebufferData, GLuint framebuffer, GLsizei numAttachments, const GLenum* attachments) DECLARE_GL_FUNCTION_END_NO_RETURN(void, InvalidateNamedFramebufferData, framebuffer, numAttachments, attachments)
+DECLARE_GL_FUNCTION_HEAD(void, InvalidateNamedFramebufferSubData, GLuint framebuffer, GLsizei numAttachments, const GLenum* attachments, GLint x, GLint y, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, InvalidateNamedFramebufferSubData, framebuffer, numAttachments, attachments, x, y, width, height)
+DECLARE_GL_FUNCTION_HEAD(void, ClearNamedFramebufferiv, GLuint framebuffer, GLenum buffer, GLint drawbuffer, const GLint* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClearNamedFramebufferiv, framebuffer, buffer, drawbuffer, value)
+DECLARE_GL_FUNCTION_HEAD(void, ClearNamedFramebufferuiv, GLuint framebuffer, GLenum buffer, GLint drawbuffer, const GLuint* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClearNamedFramebufferuiv, framebuffer, buffer, drawbuffer, value)
 DECLARE_GL_FUNCTION_HEAD(void, ClearNamedFramebufferfv, GLuint framebuffer, GLenum buffer, GLint drawbuffer, const GLfloat* value) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClearNamedFramebufferfv, framebuffer, buffer, drawbuffer, value)
 DECLARE_GL_FUNCTION_HEAD(void, ClearNamedFramebufferfi, GLuint framebuffer, GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ClearNamedFramebufferfi, framebuffer, buffer, drawbuffer, depth, stencil)
 DECLARE_GL_FUNCTION_HEAD(void, BlitNamedFramebuffer, GLuint readFramebuffer, GLuint drawFramebuffer, GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BlitNamedFramebuffer, readFramebuffer, drawFramebuffer, srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter)
 DECLARE_GL_FUNCTION_HEAD(GLenum, CheckNamedFramebufferStatus, GLuint framebuffer, GLenum target) DECLARE_GL_FUNCTION_END(GLenum, CheckNamedFramebufferStatus, framebuffer, target)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetNamedFramebufferParameteriv, GLuint framebuffer, GLenum pname, GLint* param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetNamedFramebufferParameteriv, framebuffer, pname, param)
+DECLARE_GL_FUNCTION_HEAD(void, GetNamedFramebufferParameteriv, GLuint framebuffer, GLenum pname, GLint* param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetNamedFramebufferParameteriv, framebuffer, pname, param)
 DECLARE_GL_FUNCTION_HEAD(void, GetNamedFramebufferAttachmentParameteriv, GLuint framebuffer, GLenum attachment, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetNamedFramebufferAttachmentParameteriv, framebuffer, attachment, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, CreateRenderbuffers, GLsizei n, GLuint* renderbuffers) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CreateRenderbuffers, n, renderbuffers)
 DECLARE_GL_FUNCTION_HEAD(void, NamedRenderbufferStorage, GLuint renderbuffer, GLenum internalformat, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedRenderbufferStorage, renderbuffer, internalformat, width, height)
 DECLARE_GL_FUNCTION_HEAD(void, NamedRenderbufferStorageMultisample, GLuint renderbuffer, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, NamedRenderbufferStorageMultisample, renderbuffer, samples, internalformat, width, height)
 DECLARE_GL_FUNCTION_HEAD(void, GetNamedRenderbufferParameteriv, GLuint renderbuffer, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetNamedRenderbufferParameteriv, renderbuffer, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, CreateTextures, GLenum target, GLsizei n, GLuint* textures) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CreateTextures, target, n, textures)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TextureBuffer, GLuint texture, GLenum internalformat, GLuint buffer) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TextureBuffer, texture, internalformat, buffer)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TextureBufferRange, GLuint texture, GLenum internalformat, GLuint buffer, GLintptr offset, GLsizeiptr size) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TextureBufferRange, texture, internalformat, buffer, offset, size)
+DECLARE_GL_FUNCTION_HEAD(void, TextureBuffer, GLuint texture, GLenum internalformat, GLuint buffer) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureBuffer, texture, internalformat, buffer)
+DECLARE_GL_FUNCTION_HEAD(void, TextureBufferRange, GLuint texture, GLenum internalformat, GLuint buffer, GLintptr offset, GLsizeiptr size) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureBufferRange, texture, internalformat, buffer, offset, size)
 DECLARE_GL_FUNCTION_HEAD(void, TextureStorage1D, GLuint texture, GLsizei levels, GLenum internalformat, GLsizei width) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureStorage1D, texture, levels, internalformat, width)
 DECLARE_GL_FUNCTION_HEAD(void, TextureStorage2D, GLuint texture, GLsizei levels, GLenum internalformat, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureStorage2D, texture, levels, internalformat, width, height)
 DECLARE_GL_FUNCTION_HEAD(void, TextureStorage3D, GLuint texture, GLsizei levels, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureStorage3D, texture, levels, internalformat, width, height, depth)
@@ -1059,12 +1047,12 @@ DECLARE_GL_FUNCTION_HEAD(void, TextureStorage3DMultisample, GLuint texture, GLsi
 DECLARE_GL_FUNCTION_HEAD(void, TextureSubImage1D, GLuint texture, GLint level, GLint xoffset, GLsizei width, GLenum format, GLenum type, const void* pixels) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureSubImage1D, texture, level, xoffset, width, format, type, pixels)
 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_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_STUB_HEAD(void, CopyTextureSubImage1D, GLuint texture, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CopyTextureSubImage1D, texture, level, xoffset, x, y, width)
+DECLARE_GL_FUNCTION_HEAD(void, CompressedTextureSubImage1D, GLuint texture, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const void* data) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CompressedTextureSubImage1D, texture, level, xoffset, width, 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_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_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)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, CopyTextureSubImage3D, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint x, GLint y, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CopyTextureSubImage3D, texture, level, xoffset, yoffset, zoffset, x, y, width, height)
+DECLARE_GL_FUNCTION_HEAD(void, CopyTextureSubImage3D, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint x, GLint y, GLsizei width, GLsizei height) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CopyTextureSubImage3D, texture, level, xoffset, yoffset, zoffset, x, y, width, height)
 DECLARE_GL_FUNCTION_HEAD(void, TextureParameterf, GLuint texture, GLenum pname, GLfloat param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureParameterf, texture, pname, param)
 DECLARE_GL_FUNCTION_HEAD(void, TextureParameterfv, GLuint texture, GLenum pname, const GLfloat* param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureParameterfv, texture, pname, param)
 DECLARE_GL_FUNCTION_HEAD(void, TextureParameteri, GLuint texture, GLenum pname, GLint param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureParameteri, texture, pname, param)
@@ -1074,7 +1062,7 @@ DECLARE_GL_FUNCTION_HEAD(void, TextureParameteriv, GLuint texture, GLenum pname,
 DECLARE_GL_FUNCTION_HEAD(void, GenerateTextureMipmap, GLuint texture) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GenerateTextureMipmap, texture)
 DECLARE_GL_FUNCTION_HEAD(void, BindTextureUnit, GLuint unit, GLuint texture) DECLARE_GL_FUNCTION_END_NO_RETURN(void, BindTextureUnit, unit, texture)
 DECLARE_GL_FUNCTION_HEAD(void, GetTextureImage, GLuint texture, GLint level, GLenum format, GLenum type, GLsizei bufSize, void* pixels) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTextureImage, texture, level, format, type, bufSize, pixels)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetCompressedTextureImage, GLuint texture, GLint level, GLsizei bufSize, void* pixels) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetCompressedTextureImage, texture, level, bufSize, pixels)
+DECLARE_GL_FUNCTION_HEAD(void, GetCompressedTextureImage, GLuint texture, GLint level, GLsizei bufSize, void* pixels) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetCompressedTextureImage, texture, level, bufSize, pixels)
 DECLARE_GL_FUNCTION_HEAD(void, GetTextureLevelParameterfv, GLuint texture, GLint level, GLenum pname, GLfloat* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTextureLevelParameterfv, texture, level, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetTextureLevelParameteriv, GLuint texture, GLint level, GLenum pname, GLint* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTextureLevelParameteriv, texture, level, pname, params)
 DECLARE_GL_FUNCTION_HEAD(void, GetTextureParameterfv, GLuint texture, GLenum pname, GLfloat* params) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTextureParameterfv, texture, pname, params)
@@ -1090,18 +1078,18 @@ DECLARE_GL_FUNCTION_HEAD(void, VertexArrayVertexBuffers, GLuint vaobj, GLuint fi
 DECLARE_GL_FUNCTION_HEAD(void, VertexArrayAttribBinding, GLuint vaobj, GLuint attribindex, GLuint bindingindex) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexArrayAttribBinding, vaobj, attribindex, bindingindex)
 DECLARE_GL_FUNCTION_HEAD(void, VertexArrayAttribFormat, GLuint vaobj, GLuint attribindex, GLint size, GLenum type, GLboolean normalized, GLuint relativeoffset) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexArrayAttribFormat, vaobj, attribindex, size, type, normalized, relativeoffset)
 DECLARE_GL_FUNCTION_HEAD(void, VertexArrayAttribIFormat, GLuint vaobj, GLuint attribindex, GLint size, GLenum type, GLuint relativeoffset) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexArrayAttribIFormat, vaobj, attribindex, size, type, relativeoffset)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, VertexArrayAttribLFormat, GLuint vaobj, GLuint attribindex, GLint size, GLenum type, GLuint relativeoffset) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, VertexArrayAttribLFormat, vaobj, attribindex, size, type, relativeoffset)
+DECLARE_GL_FUNCTION_HEAD(void, VertexArrayAttribLFormat, GLuint vaobj, GLuint attribindex, GLint size, GLenum type, GLuint relativeoffset) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexArrayAttribLFormat, vaobj, attribindex, size, type, relativeoffset)
 DECLARE_GL_FUNCTION_HEAD(void, VertexArrayBindingDivisor, GLuint vaobj, GLuint bindingindex, GLuint divisor) DECLARE_GL_FUNCTION_END_NO_RETURN(void, VertexArrayBindingDivisor, vaobj, bindingindex, divisor)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetVertexArrayiv, GLuint vaobj, GLenum pname, GLint* param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetVertexArrayiv, vaobj, pname, param)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetVertexArrayIndexediv, GLuint vaobj, GLuint index, GLenum pname, GLint* param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetVertexArrayIndexediv, vaobj, index, pname, param)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetVertexArrayIndexed64iv, GLuint vaobj, GLuint index, GLenum pname, GLint64* param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetVertexArrayIndexed64iv, vaobj, index, pname, param)
+DECLARE_GL_FUNCTION_HEAD(void, GetVertexArrayiv, GLuint vaobj, GLenum pname, GLint* param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetVertexArrayiv, vaobj, pname, param)
+DECLARE_GL_FUNCTION_HEAD(void, GetVertexArrayIndexediv, GLuint vaobj, GLuint index, GLenum pname, GLint* param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetVertexArrayIndexediv, vaobj, index, pname, param)
+DECLARE_GL_FUNCTION_HEAD(void, GetVertexArrayIndexed64iv, GLuint vaobj, GLuint index, GLenum pname, GLint64* param) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetVertexArrayIndexed64iv, vaobj, index, pname, param)
 DECLARE_GL_FUNCTION_HEAD(void, CreateSamplers, GLsizei n, GLuint* samplers) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CreateSamplers, n, samplers)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, CreateProgramPipelines, GLsizei n, GLuint* pipelines) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CreateProgramPipelines, n, pipelines)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, CreateQueries, GLenum target, GLsizei n, GLuint* ids) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, CreateQueries, target, n, ids)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetQueryBufferObjecti64v, GLuint id, GLuint buffer, GLenum pname, GLintptr offset) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetQueryBufferObjecti64v, id, buffer, pname, offset)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetQueryBufferObjectiv, GLuint id, GLuint buffer, GLenum pname, GLintptr offset) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetQueryBufferObjectiv, id, buffer, pname, offset)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetQueryBufferObjectui64v, GLuint id, GLuint buffer, GLenum pname, GLintptr offset) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetQueryBufferObjectui64v, id, buffer, pname, offset)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, GetQueryBufferObjectuiv, GLuint id, GLuint buffer, GLenum pname, GLintptr offset) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetQueryBufferObjectuiv, id, buffer, pname, offset)
+DECLARE_GL_FUNCTION_HEAD(void, CreateProgramPipelines, GLsizei n, GLuint* pipelines) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CreateProgramPipelines, n, pipelines)
+DECLARE_GL_FUNCTION_HEAD(void, CreateQueries, GLenum target, GLsizei n, GLuint* ids) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CreateQueries, target, n, ids)
+DECLARE_GL_FUNCTION_HEAD(void, GetQueryBufferObjecti64v, GLuint id, GLuint buffer, GLenum pname, GLintptr offset) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetQueryBufferObjecti64v, id, buffer, pname, offset)
+DECLARE_GL_FUNCTION_HEAD(void, GetQueryBufferObjectiv, GLuint id, GLuint buffer, GLenum pname, GLintptr offset) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetQueryBufferObjectiv, id, buffer, pname, offset)
+DECLARE_GL_FUNCTION_HEAD(void, GetQueryBufferObjectui64v, GLuint id, GLuint buffer, GLenum pname, GLintptr offset) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetQueryBufferObjectui64v, id, buffer, pname, offset)
+DECLARE_GL_FUNCTION_HEAD(void, GetQueryBufferObjectuiv, GLuint id, GLuint buffer, GLenum pname, GLintptr offset) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetQueryBufferObjectuiv, id, buffer, pname, offset)
 DECLARE_GL_FUNCTION_HEAD(void, GetTextureSubImage, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, GLsizei bufSize, void* pixels) DECLARE_GL_FUNCTION_END_NO_RETURN(void, GetTextureSubImage, texture, level, xoffset, yoffset, zoffset, width, height, depth, format, type, bufSize, pixels)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetCompressedTextureSubImage, GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLsizei bufSize, void* pixels) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetCompressedTextureSubImage, texture, level, xoffset, yoffset, zoffset, width, height, depth, bufSize, pixels)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnCompressedTexImage, GLenum target, GLint lod, GLsizei bufSize, void* pixels) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetnCompressedTexImage, target, lod, bufSize, pixels)
@@ -1119,11 +1107,11 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnConvolutionFilter, GLenum target, GLenum
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnSeparableFilter, GLenum target, GLenum format, GLenum type, GLsizei rowBufSize, void* row, GLsizei columnBufSize, void* column, void* span) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetnSeparableFilter, target, format, type, rowBufSize, row, columnBufSize, column, span)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnHistogram, GLenum target, GLboolean reset, GLenum format, GLenum type, GLsizei bufSize, void* values) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetnHistogram, target, reset, format, type, bufSize, values)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetnMinmax, GLenum target, GLboolean reset, GLenum format, GLenum type, GLsizei bufSize, void* values) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetnMinmax, target, reset, format, type, bufSize, values)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TextureBarrier, void) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TextureBarrier, )
-DECLARE_GL_FUNCTION_STUB_HEAD(void, SpecializeShader, GLuint shader, const GLchar* pEntryPoint, GLuint numSpecializationConstants, const GLuint* pConstantIndex, const GLuint* pConstantValue) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, SpecializeShader, shader, pEntryPoint, numSpecializationConstants, pConstantIndex, pConstantValue)
+DECLARE_GL_FUNCTION_HEAD(void, TextureBarrier, void) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureBarrier, )
+DECLARE_GL_FUNCTION_HEAD(void, SpecializeShader, GLuint shader, const GLchar* pEntryPoint, GLuint numSpecializationConstants, const GLuint* pConstantIndex, const GLuint* pConstantValue) DECLARE_GL_FUNCTION_END_NO_RETURN(void, SpecializeShader, shader, pEntryPoint, numSpecializationConstants, pConstantIndex, pConstantValue)
 DECLARE_GL_FUNCTION_HEAD(void, MultiDrawArraysIndirectCount, GLenum mode, const void* indirect, GLintptr drawcount, GLsizei maxdrawcount, GLsizei stride) DECLARE_GL_FUNCTION_END_NO_RETURN(void, MultiDrawArraysIndirectCount, mode, indirect, drawcount, maxdrawcount, stride)
 DECLARE_GL_FUNCTION_HEAD(void, MultiDrawElementsIndirectCount, GLenum mode, GLenum type, const void* indirect, GLintptr drawcount, GLsizei maxdrawcount, GLsizei stride) DECLARE_GL_FUNCTION_END_NO_RETURN(void, MultiDrawElementsIndirectCount, mode, type, indirect, drawcount, maxdrawcount, stride)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, PolygonOffsetClamp, GLfloat factor, GLfloat units, GLfloat clamp) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PolygonOffsetClamp, factor, units, clamp)
+DECLARE_GL_FUNCTION_HEAD(void, PolygonOffsetClamp, GLfloat factor, GLfloat units, GLfloat clamp) DECLARE_GL_FUNCTION_END_NO_RETURN(void, PolygonOffsetClamp, factor, units, clamp)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PrimitiveBoundingBoxARB, GLfloat minX, GLfloat minY, GLfloat minZ, GLfloat minW, GLfloat maxX, GLfloat maxY, GLfloat maxZ, GLfloat maxW) DECLARE_GL_FUNCTION_STUB_END(void, PrimitiveBoundingBoxARB, minX, minY, minZ, minW, maxX, maxY, maxZ, maxW)
 DECLARE_GL_FUNCTION_STUB_HEAD(GLuint64, GetTextureHandleARB, GLuint texture) DECLARE_GL_FUNCTION_STUB_END(GLuint64, GetTextureHandleARB, texture)
 DECLARE_GL_FUNCTION_STUB_HEAD(GLuint64, GetTextureSamplerHandleARB, GLuint texture, GLuint sampler) DECLARE_GL_FUNCTION_STUB_END(GLuint64, GetTextureSamplerHandleARB, texture, sampler)
@@ -1162,7 +1150,7 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, GetProgramLocalParameterdvARB, GLenum target
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetProgramLocalParameterfvARB, GLenum target, GLuint index, GLfloat* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetProgramLocalParameterfvARB, target, index, params)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetProgramStringARB, GLenum target, GLenum pname, void* string) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetProgramStringARB, target, pname, string)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, FramebufferTextureFaceARB, GLenum target, GLenum attachment, GLuint texture, GLint level, GLenum face) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, FramebufferTextureFaceARB, target, attachment, texture, level, face)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, SpecializeShaderARB, GLuint shader, const GLchar* pEntryPoint, GLuint numSpecializationConstants, const GLuint* pConstantIndex, const GLuint* pConstantValue) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, SpecializeShaderARB, shader, pEntryPoint, numSpecializationConstants, pConstantIndex, pConstantValue)
+DECLARE_GL_FUNCTION_HEAD(void, SpecializeShaderARB, GLuint shader, const GLchar* pEntryPoint, GLuint numSpecializationConstants, const GLuint* pConstantIndex, const GLuint* pConstantValue) DECLARE_GL_FUNCTION_END_NO_RETURN(void, SpecializeShader, shader, pEntryPoint, numSpecializationConstants, pConstantIndex, pConstantValue)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform1i64ARB, GLint location, GLint64 x) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform1i64ARB, location, x)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform2i64ARB, GLint location, GLint64 x, GLint64 y) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform2i64ARB, location, x, y)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, Uniform3i64ARB, GLint location, GLint64 x, GLint64 y, GLint64 z) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, Uniform3i64ARB, location, x, y, z)
@@ -1272,7 +1260,7 @@ 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)
@@ -1380,7 +1368,7 @@ 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)
@@ -1847,9 +1835,9 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, GetBooleanIndexedvEXT, GLenum target, GLuint
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedTextureImage3DEXT, GLuint texture, 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, CompressedTextureImage3DEXT, texture, target, level, internalformat, width, height, depth, border, imageSize, bits)
 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_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_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_END_NO_RETURN(void, CompressedTextureSubImage3D, texture, level, xoffset, yoffset, zoffset, width, height, depth, 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_HEAD(void, CompressedTextureSubImage1DEXT, GLuint texture, GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const void* bits) DECLARE_GL_FUNCTION_END_NO_RETURN(void, CompressedTextureSubImage1D, texture, 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)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, CompressedMultiTexImage2DEXT, GLenum texunit, 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, CompressedMultiTexImage2DEXT, texunit, target, level, internalformat, width, height, border, imageSize, bits)
@@ -2061,7 +2049,7 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, GetPixelTransformParameterivEXT, GLenum targ
 DECLARE_GL_FUNCTION_STUB_HEAD(void, GetPixelTransformParameterfvEXT, GLenum target, GLenum pname, GLfloat* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, GetPixelTransformParameterfvEXT, target, pname, params)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PointParameterfEXT, GLenum pname, GLfloat param) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PointParameterfEXT, pname, param)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, PointParameterfvEXT, GLenum pname, const GLfloat* params) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PointParameterfvEXT, pname, params)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, PolygonOffsetClampEXT, GLfloat factor, GLfloat units, GLfloat clamp) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, PolygonOffsetClampEXT, factor, units, clamp)
+DECLARE_GL_FUNCTION_HEAD(void, PolygonOffsetClampEXT, GLfloat factor, GLfloat units, GLfloat clamp) DECLARE_GL_FUNCTION_END_NO_RETURN(void, PolygonOffsetClamp, factor, units, clamp)
 DECLARE_GL_FUNCTION_HEAD(void, ProvokingVertexEXT, GLenum mode) DECLARE_GL_FUNCTION_END_NO_RETURN(void, ProvokingVertex, mode)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, RasterSamplesEXT, GLuint samples, GLboolean fixedsamplelocations) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, RasterSamplesEXT, samples, fixedsamplelocations)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, SecondaryColor3bEXT, GLbyte red, GLbyte green, GLbyte blue) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, SecondaryColor3bEXT, red, green, blue)
@@ -2558,7 +2546,7 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, ShadingRateImageBarrierNV, GLboolean synchro
 DECLARE_GL_FUNCTION_STUB_HEAD(void, ShadingRateImagePaletteNV, GLuint viewport, GLuint first, GLsizei count, const GLenum* rates) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ShadingRateImagePaletteNV, viewport, first, count, rates)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, ShadingRateSampleOrderNV, GLenum order) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ShadingRateSampleOrderNV, order)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, ShadingRateSampleOrderCustomNV, GLenum rate, GLuint samples, const GLint* locations) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ShadingRateSampleOrderCustomNV, rate, samples, locations)
-DECLARE_GL_FUNCTION_STUB_HEAD(void, TextureBarrierNV, void) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TextureBarrierNV, )
+DECLARE_GL_FUNCTION_HEAD(void, TextureBarrierNV, void) DECLARE_GL_FUNCTION_END_NO_RETURN(void, TextureBarrier, )
 DECLARE_GL_FUNCTION_STUB_HEAD(void, TexImage2DMultisampleCoverageNV, GLenum target, GLsizei coverageSamples, GLsizei colorSamples, GLint internalFormat, GLsizei width, GLsizei height, GLboolean fixedSampleLocations) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TexImage2DMultisampleCoverageNV, target, coverageSamples, colorSamples, internalFormat, width, height, fixedSampleLocations)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, TexImage3DMultisampleCoverageNV, GLenum target, GLsizei coverageSamples, GLsizei colorSamples, GLint internalFormat, GLsizei width, GLsizei height, GLsizei depth, GLboolean fixedSampleLocations) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TexImage3DMultisampleCoverageNV, target, coverageSamples, colorSamples, internalFormat, width, height, depth, fixedSampleLocations)
 DECLARE_GL_FUNCTION_STUB_HEAD(void, TextureImage2DMultisampleNV, GLuint texture, GLenum target, GLsizei samples, GLint internalFormat, GLsizei width, GLsizei height, GLboolean fixedSampleLocations) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, TextureImage2DMultisampleNV, texture, target, samples, internalFormat, width, height, fixedSampleLocations)
@@ -2583,7 +2571,10 @@ DECLARE_GL_FUNCTION_STUB_HEAD(void, TransformFeedbackStreamAttribsNV, GLsizei co
 DECLARE_GL_FUNCTION_STUB_HEAD(void, BindTransformFeedbackNV, GLenum target, GLuint id) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, BindTransformFeedbackNV, target, id)
 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)
-DECLARE_GL_FUNCTION_STUB_HEAD(GLboolean, IsTransformFeedbackNV, GLuint id) DECLARE_GL_FUNCTION_STUB_END(GLboolean, IsTransformFeedbackNV, id)
+MOBILEGL_GL_API GLboolean glIsTransformFeedbackNV(GLuint id) {
+    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, )
 DECLARE_GL_FUNCTION_STUB_HEAD(void, ResumeTransformFeedbackNV, void) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, ResumeTransformFeedbackNV, )
 DECLARE_GL_FUNCTION_STUB_HEAD(void, DrawTransformFeedbackNV, GLenum mode, GLuint id) DECLARE_GL_FUNCTION_STUB_END_NO_RETURN(void, DrawTransformFeedbackNV, mode, id)
@@ -3177,5 +3168,5 @@ 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.h

@@ -14,6 +14,8 @@
 namespace MobileGL::MG_Impl::GLImpl {
     /* @INSERTION_POINT:FUNCTION_DECLARATION@ */
     void ReadPixels(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, void* pixels);
+    void ReadnPixels(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, GLsizei bufSize,
+                     void* data);
     void ClearBufferfi(GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
     void ClearBufferfv(GLenum buffer, GLint drawbuffer, const GLfloat* value);
     void ClearBufferuiv(GLenum buffer, GLint drawbuffer, const GLuint* value);
@@ -56,7 +58,19 @@ namespace MobileGL::MG_Impl::GLImpl {
     void NamedFramebufferReadBuffer(GLuint framebuffer, GLenum src);
     void ClearNamedFramebufferfv(GLuint framebuffer, GLenum buffer, GLint drawbuffer, const GLfloat* value);
     void ClearNamedFramebufferfi(GLuint framebuffer, GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
+    void InvalidateNamedFramebufferData(GLuint framebuffer, GLsizei numAttachments, const GLenum* attachments);
+    void InvalidateNamedFramebufferSubData(GLuint framebuffer, GLsizei numAttachments, const GLenum* attachments,
+                                           GLint x, GLint y, GLsizei width, GLsizei height);
+    void InvalidateFramebuffer(GLenum target, GLsizei numAttachments, const GLenum* attachments);
+    void InvalidateSubFramebuffer(GLenum target, GLsizei numAttachments, const GLenum* attachments, GLint x, GLint y,
+                                  GLsizei width, GLsizei height);
+    void ClearNamedFramebufferiv(GLuint framebuffer, GLenum buffer, GLint drawbuffer, const GLint* value);
+    void ClearNamedFramebufferuiv(GLuint framebuffer, GLenum buffer, GLint drawbuffer, const GLuint* value);
     GLenum CheckNamedFramebufferStatus(GLuint framebuffer, GLenum target);
+    void GetFramebufferParameteriv(GLenum target, GLenum pname, GLint* params);
+    void FramebufferParameteri(GLenum target, GLenum pname, GLint param);
+    void GetNamedFramebufferParameteriv(GLuint framebuffer, GLenum pname, GLint* params);
+    void NamedFramebufferParameteri(GLuint framebuffer, GLenum pname, GLint param);
     void GetNamedFramebufferAttachmentParameteriv(GLuint framebuffer, GLenum attachment, GLenum pname, GLint* params);
     void BlitNamedFramebuffer(GLuint readFramebuffer, GLuint drawFramebuffer, GLint srcX0, GLint srcY0, GLint srcX1,
                               GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask,
@@ -78,6 +92,6 @@ namespace MobileGL::MG_Impl::GLImpl {
             SharedPtr<MG_State::GLState::ITextureObject> stencilAttachment;
         };
 
-        extern UniquePtr<DefaultFramebufferInfo> pDefaultFramebufferInfo;
+        extern UniquePtr<DefaultFramebufferInfo>& pDefaultFramebufferInfo;
     } // namespace FramebufferImpl
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Framebuffer/Validators.cpp

@@ -7,6 +7,7 @@
 // End of Source File Header
 
 #include "Validators.h"
+#include <MG_Backend/BackendObjects.h>
 #include <MG_State/GLState/Core.h>
 #include <MG_State/GLState/ErrorState/Error.h>
 #include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
@@ -60,6 +61,26 @@ namespace MobileGL::MG_Impl::GLImpl::FramebufferImpl {
         return true;
     }
 
+    Bool ValidateColorAttachmentInRange(FramebufferAttachmentType attachment, const char* caller) {
+        const auto first = static_cast<SizeT>(FramebufferAttachmentType::Color0);
+        const auto index = static_cast<SizeT>(attachment);
+        if (index < first) return true;
+        const auto colorIndex = index - first;
+        const auto limit = static_cast<SizeT>(
+            MG_Backend::pActiveBackendObject ? MG_Backend::pActiveBackendObject->GetDynamicParameters()
+                                                   .MaxColorAttachments
+                                             : static_cast<Int>(MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS));
+        if (colorIndex >= limit) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl/FramebufferImpl", caller,
+                    std::format("Colour attachment {} is beyond GL_MAX_COLOR_ATTACHMENTS ({}).", colorIndex, limit)));
+            return false;
+        }
+        return true;
+    }
+
     Bool ValidateRenderbufferTarget(RenderbufferTarget target) {
         if (target == RenderbufferTarget::Unknown) {
             using namespace MG_Util;
@@ -97,4 +118,100 @@ namespace MobileGL::MG_Impl::GLImpl::FramebufferImpl {
                                          std::format("Renderbuffer name {} is not valid.", index)));
         return false;
     }
+
+    Bool ValidateFramebufferParameterPname(GLenum pname, Bool isDefaultFramebuffer, Bool forSetter,
+                                           const char* caller) {
+        Bool isDefaultParameter = false;
+        switch (pname) {
+        case GL_FRAMEBUFFER_DEFAULT_WIDTH:
+        case GL_FRAMEBUFFER_DEFAULT_HEIGHT:
+        case GL_FRAMEBUFFER_DEFAULT_LAYERS:
+        case GL_FRAMEBUFFER_DEFAULT_SAMPLES:
+        case GL_FRAMEBUFFER_DEFAULT_FIXED_SAMPLE_LOCATIONS:
+            isDefaultParameter = true;
+            break;
+        case GL_DOUBLEBUFFER:
+        case GL_IMPLEMENTATION_COLOR_READ_FORMAT:
+        case GL_IMPLEMENTATION_COLOR_READ_TYPE:
+        case GL_SAMPLES:
+        case GL_SAMPLE_BUFFERS:
+        case GL_STEREO:
+            // Queryable only; glFramebufferParameteri sets none of these.
+            if (forSetter) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidEnum,
+                    MakeUnique<GenericErrorInfo>(
+                        "MG_Impl/GLImpl/FramebufferImpl", caller,
+                        std::format("pname {} is not settable on a framebuffer.",
+                                    MG_Util::ConvertGLEnumToString(pname))));
+                return false;
+            }
+            break;
+        default:
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl/FramebufferImpl", caller,
+                    std::format("pname {} is not a framebuffer parameter.",
+                                MG_Util::ConvertGLEnumToString(pname))));
+            return false;
+        }
+
+        // The default framebuffer has no DEFAULT_* state of its own - its shape comes from the
+        // surface - so those names are accepted enums it simply cannot answer or accept.
+        if (isDefaultFramebuffer && isDefaultParameter) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl/FramebufferImpl", caller,
+                    std::format("pname {} does not apply to the default framebuffer.",
+                                MG_Util::ConvertGLEnumToString(pname))));
+            return false;
+        }
+        return true;
+    }
+
+    Bool ValidateReadFramebufferForCopy(const char* caller) {
+        auto& framebufferObject =
+            MG_State::pGLContext->GetFramebufferBindingSlot(FramebufferTarget::Read).GetBoundObject();
+        if (!framebufferObject || !framebufferObject->CheckCompleteness()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidFramebufferOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl/FramebufferImpl", caller,
+                                             "Read framebuffer is not framebuffer complete."));
+            return false;
+        }
+
+        const FramebufferAttachmentType readBuffer = framebufferObject->GetReadBuffer();
+        if (readBuffer == FramebufferAttachmentType::None ||
+            !framebufferObject->GetAttachment(readBuffer).IsValid()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl/FramebufferImpl", caller,
+                                             "Read buffer names no attachment of the read framebuffer."));
+            return false;
+        }
+
+        // SAMPLE_BUFFERS is one whenever the read buffer resolves to multisample storage. A
+        // multisample texture says so by its target - its sample count can legally be one - while a
+        // renderbuffer says so by having been given a non-zero sample count.
+        const auto& readAttachment = framebufferObject->GetAttachment(readBuffer);
+        Bool isMultisampled = false;
+        if (readAttachment.IsRenderbuffer() && readAttachment.GetRenderbuffer()) {
+            isMultisampled = readAttachment.GetRenderbuffer()->GetSamples() > 0;
+        } else if (readAttachment.IsTexture() && readAttachment.GetTexture()) {
+            const auto target = readAttachment.GetTexture()->GetTarget();
+            isMultisampled = target == TextureTarget::Texture2DMultisample ||
+                             target == TextureTarget::Texture2DMultisampleArray;
+        }
+        if (isMultisampled) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl/FramebufferImpl", caller,
+                                             "Cannot copy from a multisampled read framebuffer."));
+            return false;
+        }
+
+        return true;
+    }
 } // namespace MobileGL::MG_Impl::GLImpl::FramebufferImpl

MobileGL/MG_Impl/GLImpl/Framebuffer/Validators.h

@@ -14,6 +14,21 @@ namespace MobileGL::MG_Impl::GLImpl::FramebufferImpl {
     Bool ValidateFramebufferTarget(FramebufferTarget target);
     Bool ValidateFramebufferName(Uint index, Bool allowZero = true);
     Bool ValidateFramebufferAttachmentType(FramebufferAttachmentType attachment);
+    // GL_COLOR_ATTACHMENTn is a token per n up to 31, but only the first GL_MAX_COLOR_ATTACHMENTS of
+    // them name an attachment point of a framebuffer object; the rest are INVALID_OPERATION for the
+    // attaching entry points (GL 4.6 core 9.2.7). Non-colour attachments pass through unchanged.
+    Bool ValidateColorAttachmentInRange(FramebufferAttachmentType attachment, const char* caller);
     Bool ValidateRenderbufferTarget(RenderbufferTarget target);
     Bool ValidateRenderbufferName(Uint index, Bool allowZero = true);
+    // The read-framebuffer preconditions the CopyTexSubImage family shares (GL 4.6 core 8.6): the
+    // read framebuffer must be complete, its read buffer must name a real attachment, and it must
+    // not be multisampled. Incompleteness is INVALID_FRAMEBUFFER_OPERATION, the other two are
+    // INVALID_OPERATION.
+    Bool ValidateReadFramebufferForCopy(const char* caller);
+    // The pname sets of glGet/FramebufferParameteri (GL 4.6 core 9.2.3). Order matters and is part
+    // of the contract: a name outside the table is INVALID_ENUM, and only then is a name that the
+    // DEFAULT framebuffer does not answer INVALID_OPERATION. Testing the framebuffer kind first
+    // would turn GL_FRAMEBUFFER_DEFAULT_WIDTH on framebuffer zero into the wrong error.
+    Bool ValidateFramebufferParameterPname(GLenum pname, Bool isDefaultFramebuffer, Bool forSetter,
+                                           const char* caller);
 } // namespace MobileGL::MG_Impl::GLImpl::FramebufferImpl

MobileGL/MG_Impl/GLImpl/Getter/GL_Getter.h

@@ -19,6 +19,30 @@ 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();
+    // The GL_MAX_SAMPLES value MobileGL advertises, i.e. the driver's value floored to the GL
+    // core minimum of 4. This is the RENDERBUFFER ceiling; the three per-category texture
+    // ceilings below have a minimum of one and are reported as probed.
+    GLint GetAdvertisedMaxSamples();
+    // Exactly what GL_MAX_COLOR_TEXTURE_SAMPLES / GL_MAX_DEPTH_TEXTURE_SAMPLES /
+    // GL_MAX_INTEGER_SAMPLES report: the probed backend limit floored at the GL 4.6 core minimum
+    // of ONE (table 23.53). Exported so the frontend's storage validation enforces exactly what
+    // the query promised - it used to floor both at 4 and then let the backend quietly
+    // under-allocate whatever the driver could not actually provide.
+    GLint GetAdvertisedColorTextureMaxSamples();
+    GLint GetAdvertisedDepthTextureMaxSamples();
+    GLint GetAdvertisedIntegerMaxSamples();
+    // What glGetIntegerv(GL_SAMPLES) answers for the CURRENT draw framebuffer: the largest sample
+    // count over its attachments, and 0 for a single-sample or default framebuffer (GL 4.6 core
+    // 9.2.3 / 22.2 - GL_SAMPLE_BUFFERS is 1 exactly when this is non-zero).
+    //
+    // Shared rather than duplicated because two callers need the identical number and disagreeing
+    // would be a silent bug: the query itself, and the draw path's write of the reserved
+    // gl_NumSamples stand-in - a shader comparing gl_NumSamples against glGetIntegerv(GL_SAMPLES)
+    // is exactly what the sample_variables CTS does.
+    GLint ResolveDrawFramebufferSampleCount();
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Program/GL_Program.h

@@ -13,6 +13,12 @@ namespace MobileGL::MG_Impl::GLImpl {
     void AttachShader(GLuint program, GLuint shader);
     void BindAttribLocation(GLuint program, GLuint index, const GLchar* name);
     void CompileShader(GLuint shader);
+    // GL_ARB_gl_spirv, core since 4.6. The pair is a two-step operation: glShaderBinary attaches
+    // the module to one or more shader objects, glSpecializeShader names its entry point and
+    // supplies its specialization constants and is what actually compiles them.
+    void ShaderBinary(GLsizei count, const GLuint* shaders, GLenum binaryformat, const void* binary, GLsizei length);
+    void SpecializeShader(GLuint shader, const GLchar* pEntryPoint, GLuint numSpecializationConstants,
+                          const GLuint* pConstantIndex, const GLuint* pConstantValue);
     GLuint CreateProgram(void);
     GLuint CreateShader(GLenum type);
     void DeleteProgram(GLuint program);
@@ -42,6 +48,10 @@ 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);
@@ -136,6 +146,49 @@ namespace MobileGL::MG_Impl::GLImpl {
                               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 GetActiveAtomicCounterBufferiv(GLuint program, GLuint bufferIndex, GLenum pname, GLint* params);
     void ShaderStorageBlockBinding(GLuint program, GLuint storageBlockIndex, GLuint storageBlockBinding);
+    void Uniform1d(GLint location, GLdouble v0);
+    void Uniform1dv(GLint location, GLsizei count, const GLdouble* value);
+    void ProgramUniform1d(GLuint program, GLint location, GLdouble v0);
+    void ProgramUniform1dv(GLuint program, GLint location, GLsizei count, const GLdouble* value);
+    void Uniform2d(GLint location, GLdouble v0, GLdouble v1);
+    void Uniform2dv(GLint location, GLsizei count, const GLdouble* value);
+    void ProgramUniform2d(GLuint program, GLint location, GLdouble v0, GLdouble v1);
+    void ProgramUniform2dv(GLuint program, GLint location, GLsizei count, const GLdouble* value);
+    void Uniform3d(GLint location, GLdouble v0, GLdouble v1, GLdouble v2);
+    void Uniform3dv(GLint location, GLsizei count, const GLdouble* value);
+    void ProgramUniform3d(GLuint program, GLint location, GLdouble v0, GLdouble v1, GLdouble v2);
+    void ProgramUniform3dv(GLuint program, GLint location, GLsizei count, const GLdouble* value);
+    void Uniform4d(GLint location, GLdouble v0, GLdouble v1, GLdouble v2, GLdouble v3);
+    void Uniform4dv(GLint location, GLsizei count, const GLdouble* value);
+    void ProgramUniform4d(GLuint program, GLint location, GLdouble v0, GLdouble v1, GLdouble v2, GLdouble v3);
+    void ProgramUniform4dv(GLuint program, GLint location, GLsizei count, const GLdouble* value);
+    void UniformMatrix2dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix2dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void UniformMatrix3dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix3dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void UniformMatrix4dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix4dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void UniformMatrix2x3dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix2x3dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void UniformMatrix2x4dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix2x4dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void UniformMatrix3x2dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix3x2dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void UniformMatrix3x4dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix3x4dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void UniformMatrix4x2dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix4x2dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void UniformMatrix4x3dv(GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    void ProgramUniformMatrix4x3dv(GLuint program, GLint location, GLsizei count, GLboolean transpose, const GLdouble* value);
+    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);
+    void GetTransformFeedbackVarying(GLuint program, GLuint index, GLsizei bufSize, GLsizei* length, GLsizei* size,
+                                     GLenum* type, GLchar* name);
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Program/GL_ProgramPipeline.cpp

@@ -0,0 +1,250 @@
+// MobileGL - MobileGL/MG_Impl/GLImpl/Program/GL_ProgramPipeline.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 "GL_ProgramPipeline.h"
+
+#include <MG_State/GLState/Core.h>
+#include <MG_State/GLState/ErrorState/ErrorInfo.h>
+#include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
+
+namespace MobileGL::MG_Impl::GLImpl {
+    namespace {
+        void RecordPipelineError(ErrorCode code, const char* function, String message) {
+            MG_State::pGLContext->RecordError(
+                code, MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", function, Move(message)));
+        }
+
+        // GL 4.6 core 7.4 asks only that the name came from GenProgramPipelines and has not been
+        // 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) {
+            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 &object;
+        }
+
+        Bool ValidatePipelineCount(GLsizei n, const char* function) {
+            if (n < 0) {
+                RecordPipelineError(ErrorCode::InvalidValue, function, "n must be non-negative.");
+                return false;
+            }
+            return true;
+        }
+
+        // GL 4.6 core table 7.1 maps each stage bit onto a shader stage.
+        Bool TryResolveStageBit(GLbitfield bit, ShaderStage& outStage) {
+            switch (bit) {
+            case GL_VERTEX_SHADER_BIT: outStage = ShaderStage::Vertex; return true;
+            case GL_TESS_CONTROL_SHADER_BIT: outStage = ShaderStage::TessControl; return true;
+            case GL_TESS_EVALUATION_SHADER_BIT: outStage = ShaderStage::TessEval; return true;
+            case GL_GEOMETRY_SHADER_BIT: outStage = ShaderStage::Geometry; return true;
+            case GL_FRAGMENT_SHADER_BIT: outStage = ShaderStage::Fragment; return true;
+            case GL_COMPUTE_SHADER_BIT: outStage = ShaderStage::Compute; return true;
+            default: return false;
+            }
+        }
+
+        constexpr GLbitfield kAllStageBits = GL_VERTEX_SHADER_BIT | GL_TESS_CONTROL_SHADER_BIT |
+                                            GL_TESS_EVALUATION_SHADER_BIT | GL_GEOMETRY_SHADER_BIT |
+                                            GL_FRAGMENT_SHADER_BIT | GL_COMPUTE_SHADER_BIT;
+    } // namespace
+
+    void GenProgramPipelines(GLsizei n, GLuint* pipelines) {
+        if (!ValidatePipelineCount(n, __func__)) return;
+        if (n == 0 || !pipelines) return;
+
+        static thread_local Vector<GLuint> names;
+        MG_State::pGLContext->GenProgramPipelineNames(static_cast<Uint>(n), names);
+        Memcpy(pipelines, names.data(), static_cast<SizeT>(n) * sizeof(GLuint));
+    }
+
+    void CreateProgramPipelines(GLsizei n, GLuint* pipelines) {
+        if (!ValidatePipelineCount(n, __func__)) return;
+        if (n == 0 || !pipelines) return;
+
+        static thread_local Vector<GLuint> names;
+        MG_State::pGLContext->GenProgramPipelineNames(static_cast<Uint>(n), names);
+        for (GLsizei i = 0; i < n; ++i) {
+            pipelines[i] = names[static_cast<SizeT>(i)];
+            MG_State::pGLContext->CreateProgramPipelineObject(names[static_cast<SizeT>(i)]);
+        }
+    }
+
+    void DeleteProgramPipelines(GLsizei n, const GLuint* pipelines) {
+        if (!ValidatePipelineCount(n, __func__)) return;
+        if (!pipelines) return;
+
+        for (GLsizei i = 0; i < n; ++i) {
+            // Deleting zero, an unknown name, or a name that was only reserved is silently ignored.
+            MG_State::pGLContext->MarkProgramPipelineForDeletion(pipelines[i]);
+        }
+    }
+
+    void BindProgramPipeline(GLuint pipeline) {
+        if (pipeline != 0 && !MG_State::pGLContext->ValidateProgramPipelineName(pipeline)) {
+            RecordPipelineError(ErrorCode::InvalidOperation, __func__,
+                                std::format("Program pipeline name {} is not valid.", pipeline));
+            return;
+        }
+        MG_State::pGLContext->BindProgramPipelineObject(pipeline);
+    }
+
+    GLboolean IsProgramPipeline(GLuint pipeline) {
+        return MG_State::pGLContext->IsProgramPipelineObject(pipeline) ? GL_TRUE : GL_FALSE;
+    }
+
+    void GetProgramPipelineiv(GLuint pipeline, GLenum pname, GLint* params) {
+        const auto* pipelineObject = TryGetPipeline(pipeline, __func__);
+        if (!pipelineObject || !params) return;
+
+        const auto stageProgramName = [&](ShaderStage stage) -> GLint {
+            const auto& program = (*pipelineObject)->GetStageProgram(stage);
+            return program ? static_cast<GLint>(program->GetExternalIndex()) : 0;
+        };
+
+        switch (pname) {
+        case GL_ACTIVE_PROGRAM: {
+            const auto& active = (*pipelineObject)->GetActiveProgram();
+            *params = active ? static_cast<GLint>(active->GetExternalIndex()) : 0;
+            break;
+        }
+        case GL_VERTEX_SHADER: *params = stageProgramName(ShaderStage::Vertex); break;
+        case GL_TESS_CONTROL_SHADER: *params = stageProgramName(ShaderStage::TessControl); break;
+        case GL_TESS_EVALUATION_SHADER: *params = stageProgramName(ShaderStage::TessEval); break;
+        case GL_GEOMETRY_SHADER: *params = stageProgramName(ShaderStage::Geometry); break;
+        case GL_FRAGMENT_SHADER: *params = stageProgramName(ShaderStage::Fragment); break;
+        case GL_COMPUTE_SHADER: *params = stageProgramName(ShaderStage::Compute); break;
+        case GL_VALIDATE_STATUS: *params = (*pipelineObject)->GetValidateStatus() ? GL_TRUE : GL_FALSE; break;
+        case GL_INFO_LOG_LENGTH: {
+            // GL counts the null terminator, and reports 0 rather than 1 for an empty log.
+            const auto& log = (*pipelineObject)->GetInfoLog();
+            *params = log.empty() ? 0 : static_cast<GLint>(log.length()) + 1;
+            break;
+        }
+        default:
+            RecordPipelineError(ErrorCode::InvalidEnum, __func__,
+                                std::format("pname {} is not a program pipeline parameter.",
+                                            MG_Util::ConvertGLEnumToString(pname)));
+            break;
+        }
+    }
+
+    void GetProgramPipelineInfoLog(GLuint pipeline, GLsizei bufSize, GLsizei* length, GLchar* infoLog) {
+        const auto* pipelineObject = TryGetPipeline(pipeline, __func__);
+        if (!pipelineObject) return;
+        if (bufSize < 0) {
+            RecordPipelineError(ErrorCode::InvalidValue, __func__, "bufSize must be non-negative.");
+            return;
+        }
+        if (bufSize == 0 || !infoLog) {
+            if (length) *length = 0;
+            return;
+        }
+
+        const auto& log = (*pipelineObject)->GetInfoLog();
+        const auto copied = std::min<GLsizei>(bufSize - 1, static_cast<GLsizei>(log.length()));
+        if (copied > 0) Memcpy(infoLog, log.data(), static_cast<SizeT>(copied));
+        infoLog[copied] = '\0';
+        if (length) *length = copied;
+    }
+
+    void UseProgramStages(GLuint pipeline, GLbitfield stages, GLuint program) {
+        if (stages != GL_ALL_SHADER_BITS && (stages & ~kAllStageBits) != 0) {
+            RecordPipelineError(ErrorCode::InvalidValue, __func__, "stages names a bit that is not a shader stage.");
+            return;
+        }
+        const auto* pipelineObject = TryGetPipeline(pipeline, __func__);
+        if (!pipelineObject) return;
+
+        SharedPtr<MG_State::GLState::ProgramObject> programObject;
+        if (program != 0) {
+            if (!MG_State::pGLContext->ValidateProgramName(program)) {
+                RecordPipelineError(ErrorCode::InvalidValue, __func__,
+                                    std::format("{} is not the name of a program object.", program));
+                return;
+            }
+            programObject = MG_State::pGLContext->GetProgramObject(program);
+            if (!programObject) {
+                RecordPipelineError(ErrorCode::InvalidValue, __func__,
+                                    std::format("{} is not the name of a program object.", program));
+                return;
+            }
+            if (!programObject->GetLinkStatus()) {
+                RecordPipelineError(ErrorCode::InvalidOperation, __func__,
+                                    std::format("Program {} has not been linked successfully.", program));
+                return;
+            }
+            // GL 4.6 core 7.4: "INVALID_OPERATION is generated if program was not linked with its
+            // PROGRAM_SEPARABLE status set". The LATCHED flag is the one that decides - a program
+            // whose live flag was cleared after a separable link is still a legal stage, and a
+            // program whose live flag was set after a non-separable link is not.
+            if (!programObject->GetLinkedSeparable()) {
+                RecordPipelineError(ErrorCode::InvalidOperation, __func__,
+                                    std::format("Program {} was not linked as a separable program.", program));
+                return;
+            }
+        }
+
+        const GLbitfield selected = stages == GL_ALL_SHADER_BITS ? kAllStageBits : stages;
+        for (GLbitfield bit = 1; bit != 0 && bit <= kAllStageBits; bit <<= 1) {
+            if ((selected & bit) == 0) continue;
+            ShaderStage stage = ShaderStage::Unknown;
+            if (!TryResolveStageBit(bit, stage)) continue;
+            // program == 0 clears the stage, which is what a null program reference means here.
+            (*pipelineObject)->SetStageProgram(stage, programObject);
+        }
+    }
+
+    void ActiveShaderProgram(GLuint pipeline, GLuint program) {
+        const auto* pipelineObject = TryGetPipeline(pipeline, __func__);
+        if (!pipelineObject) return;
+
+        if (program == 0) {
+            (*pipelineObject)->SetActiveProgram(nullptr);
+            return;
+        }
+        if (!MG_State::pGLContext->ValidateProgramName(program)) {
+            RecordPipelineError(ErrorCode::InvalidValue, __func__,
+                                std::format("{} is not the name of a program object.", program));
+            return;
+        }
+        auto programObject = MG_State::pGLContext->GetProgramObject(program);
+        if (!programObject) {
+            RecordPipelineError(ErrorCode::InvalidValue, __func__,
+                                std::format("{} is not the name of a program object.", program));
+            return;
+        }
+        if (!programObject->GetLinkStatus()) {
+            RecordPipelineError(ErrorCode::InvalidOperation, __func__,
+                                std::format("Program {} has not been linked successfully.", program));
+            return;
+        }
+        (*pipelineObject)->SetActiveProgram(programObject);
+    }
+
+    void ValidateProgramPipeline(GLuint pipeline) {
+        const auto* pipelineObject = TryGetPipeline(pipeline, __func__);
+        if (!pipelineObject) return;
+        // Nothing here can fail today: MobileGL links each stage program on its own, so there is no
+        // cross-stage interface to re-check at validation time. The log stays empty, which GL allows.
+        (*pipelineObject)->SetValidateStatus(true);
+    }
+} // namespace MobileGL::MG_Impl::GLImpl