[Fix] (DirectGLES): lower gl_ClipDistance for Adreno's ESSL compiler - shadow the builtin in a Private array with constant-index flushes before EmitVertex/return, loop-copy gl_in clip distances through dynamic indices (whole-array reads segfault the Qualcomm compiler, constant-index element reads miscompile), strip the SPIRV-Cross redeclaration Adreno rejects, and split const struct-array LUT initializers so they stay dynamically indexable; quirk-gated to Qualcomm with MOBILEGL_QUIRK_CLIP_DISTANCE override

- SyncCurrentFBO skips the READ-target pass when the same GL FBO is bound as
  both draw and read (the common GL_FRAMEBUFFER case), but the read buffer
  (glReadBuffer) is only applied inside SyncToBackend's READ path — so the skip
  silently dropped every glReadBuffer change, leaving the backend read buffer
  stuck at COLOR_ATTACHMENT0.
- Extract the read-buffer application into BackendFramebufferObject::
  SyncReadBufferToBackend and invoke it from the skip branch (target == Read)
  as well as from SyncToBackend, so reads always target the right attachment.
- Bind the backend FBO as READ inside the helper before glReadBuffer, since the
  skip path only bound it as DRAW.
- Fixes KHR-GL3x.draw_buffers.draw_buffers_1 (reading COLOR_ATTACHMENT1 while
  the FBO stays GL_FRAMEBUFFER-bound returned attachment 0's value); the render
  was already correct, only the readback resolved the wrong attachment.

.gitattributes

@@ -0,0 +1,4 @@
+tools/trace_replay/fixtures/*.tgz filter=lfs diff=lfs merge=lfs -text
+tools/trace_replay/fixtures/*.png filter=lfs diff=lfs merge=lfs -text
+tools/trace_replay/fixtures/openra.tgz -filter -diff -merge -text
+tools/trace_replay/fixtures/openra.0000031249.png -filter -diff -merge -text

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

@@ -0,0 +1,239 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+if [ "$#" -lt 1 ] || [ "$#" -gt 2 ]; then
+  echo "usage: $0 <trace-case> [fixture-dir]" >&2
+  exit 2
+fi
+
+case_name="$1"
+fixture_dir="${2:-tools/trace_replay/fixtures}"
+python_bin="${PYTHON:-python3}"
+# Fixture mirrors, tried in order before falling back to Git LFS. Override the
+# whole list with MOBILEGL_TRACE_FIXTURE_MIRROR_BASES (whitespace separated);
+# MOBILEGL_TRACE_FIXTURE_MIRROR_BASE still works and is tried first.
+default_mirror_bases=(
+  "https://git.hit.moe/swung0x48/MobileGL/media/branch/dev/tools/trace_replay/fixtures"
+  "https://repo.miawa.cn/mgl/tools/trace_replay/fixtures"
+)
+if [ -n "${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}"
+
+if ! command -v "${python_bin}" >/dev/null 2>&1 && command -v python >/dev/null 2>&1; then
+  python_bin=python
+fi
+
+if ! [[ "${download_attempts}" =~ ^[1-9][0-9]*$ ]]; then
+  echo "MOBILEGL_TRACE_FIXTURE_DOWNLOAD_ATTEMPTS must be a positive integer: ${download_attempts}" >&2
+  exit 2
+fi
+if ! [[ "${retry_delay}" =~ ^[0-9]+$ ]]; then
+  echo "MOBILEGL_TRACE_FIXTURE_RETRY_DELAY must be a non-negative integer: ${retry_delay}" >&2
+  exit 2
+fi
+
+fixture_list="$("${python_bin}" tools/trace_replay/trace_cases.py \
+  --format fixture-files \
+  --case "${case_name}" \
+  --fixture-root "${fixture_dir}")"
+# Strip CR so the script also works when python emits CRLF (Git Bash on Windows).
+mapfile -t files < <(printf '%s\n' "${fixture_list}" | tr -d '\r')
+
+include="$(IFS=,; echo "${files[*]}")"
+if [ "${case_name}" = "OpenRA" ]; then
+  echo "Fixture files for ${case_name} are stored in Git: ${include}"
+  for file in "${files[@]}"; do
+    test -s "${file}"
+    if head -n 1 "${file}" | grep -q "version https://git-lfs.github.com/spec/v1"; then
+      echo "fixture should not be stored as an LFS pointer: ${file}" >&2
+      exit 1
+    fi
+  done
+  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="$2"
+  local metadata
+  local expected_oid
+  local expected_size
+  local tmp_file="${file}.tmp"
+  local attempt
+  local partial_size
+  local curl_status
+  local curl_auth
+
+  metadata="$(get_lfs_metadata "${file}")" || return 1
+  read -r expected_oid expected_size <<< "${metadata}"
+
+  if [ -f "${tmp_file}" ]; then
+    partial_size="$(wc -c < "${tmp_file}" | tr -d '[:space:]')"
+    if [ "${partial_size}" -gt "${expected_size}" ]; then
+      echo "Discarding oversized partial fixture ${tmp_file}: ${partial_size} > ${expected_size}" >&2
+      rm -f "${tmp_file}"
+    elif [ "${partial_size}" = "${expected_size}" ]; then
+      if verify_fixture_file "${tmp_file}" "${file}" "${expected_oid}" "${expected_size}"; then
+        mv "${tmp_file}" "${file}"
+        return 0
+      fi
+      rm -f "${tmp_file}"
+    fi
+  fi
+
+  for ((attempt = 1; attempt <= download_attempts; attempt++)); do
+    partial_size=0
+    if [ -f "${tmp_file}" ]; then
+      partial_size="$(wc -c < "${tmp_file}" | tr -d '[:space:]')"
+    fi
+
+    if [ "${partial_size}" -gt 0 ]; then
+      echo "Resuming mirror download for ${file} at byte ${partial_size} (attempt ${attempt}/${download_attempts})"
+    else
+      echo "Starting mirror download for ${file} (attempt ${attempt}/${download_attempts})"
+    fi
+
+    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
+      fi
+      echo "Mirror download failed integrity verification; retrying from the beginning: ${file}" >&2
+      rm -f "${tmp_file}"
+    else
+      curl_status=$?
+      partial_size=0
+      if [ -f "${tmp_file}" ]; then
+        partial_size="$(wc -c < "${tmp_file}" | tr -d '[:space:]')"
+      fi
+
+      if [ "${partial_size}" = "${expected_size}" ]; then
+        if verify_fixture_file "${tmp_file}" "${file}" "${expected_oid}" "${expected_size}"; then
+          mv "${tmp_file}" "${file}"
+          return 0
+        fi
+        rm -f "${tmp_file}"
+        partial_size=0
+      elif [ "${partial_size}" -gt "${expected_size}" ]; then
+        echo "Discarding oversized partial fixture ${tmp_file}: ${partial_size} > ${expected_size}" >&2
+        rm -f "${tmp_file}"
+        partial_size=0
+      elif [ "${curl_status}" -eq 33 ]; then
+        echo "Mirror refused the resume request; retrying from the beginning: ${file}" >&2
+        rm -f "${tmp_file}"
+        partial_size=0
+      fi
+
+      echo "Mirror download attempt ${attempt}/${download_attempts} failed with curl exit ${curl_status}; retained ${partial_size} bytes for resume: ${file}" >&2
+    fi
+
+    if [ "${attempt}" -lt "${download_attempts}" ]; then
+      sleep "${retry_delay}"
+    fi
+  done
+
+  rm -f "${tmp_file}"
+  return 1
+}
+
+fetch_from_mirror() {
+  mkdir -p "${fixture_dir}"
+  for file in "${files[@]}"; do
+    local name
+    local url
+    local base
+    local fetched=0
+    name="$(basename "${file}")"
+    for base in "${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
+      return 1
+    fi
+  done
+}
+
+if fetch_from_mirror; then
+  echo "Fetched trace fixture files for ${case_name} from mirror: ${include}"
+else
+  echo "All mirrors failed for ${case_name}; falling back to Git LFS: ${include}"
+  git lfs install --local
+  git lfs pull --include="${include}" --exclude=""
+fi
+
+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

.github/scripts/validate-plugin-apks.sh

@@ -0,0 +1,75 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+if [[ $# -ne 3 ]]; then
+  echo "Usage: $0 <aapt2> <plugin-apk> <trace-apk>" >&2
+  exit 64
+fi
+
+aapt2=$1
+plugin_apk=$2
+trace_apk=$3
+
+require() {
+  local needle=$1
+  local content=$2
+  local description=$3
+  if ! grep -Fq -- "$needle" <<<"$content"; then
+    echo "::error::Missing ${description}: ${needle}" >&2
+    exit 1
+  fi
+}
+
+for apk in "$plugin_apk" "$trace_apk"; do
+  [[ -f "$apk" ]] || { echo "::error::APK not found: $apk" >&2; exit 1; }
+done
+
+plugin_manifest=$("$aapt2" dump xmltree --file AndroidManifest.xml "$plugin_apk")
+plugin_resources=$("$aapt2" dump resources "$plugin_apk")
+plugin_resource_text=$(tr -d '"' <<<"$plugin_resources")
+trace_manifest=$("$aapt2" dump xmltree --file AndroidManifest.xml "$trace_apk")
+plugin_contents=$(unzip -Z1 "$plugin_apk")
+
+require 'top.mobilegl.plugin' "$plugin_manifest" 'plugin package name'
+require 'MobileGL' "$plugin_manifest" 'plugin label'
+require 'fclPlugin' "$plugin_manifest" 'legacy plugin marker'
+require 'fclPlugin_V2' "$plugin_manifest" 'V2 plugin marker'
+require 'LIBGL_ES=3:POJAV_RENDERER=opengles3:MOBILEGL_BACKEND_TYPE=DirectGLES' "$plugin_manifest" 'V1 DirectGLES fallback'
+require 'string/config' "$plugin_resources" 'V2 renderer configuration resource'
+require '{displayName:MobileGL,rendererId:opengles3' "$plugin_resource_text" 'V2 MobileGL entry and renderer ID'
+require 'rendererGLPath:**|libMobileGL.so' "$plugin_resource_text" 'V2 GL library path'
+require 'rendererEGLPath:**|libMobileGL.so' "$plugin_resource_text" 'V2 EGL library path'
+require 'key:LIBGL_ES,value:3' "$plugin_resource_text" 'V2 fixed LIBGL_ES variable'
+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' "$plugin_resource_text" 'V2 Vulkan subgroup toggle'
+require 'key:MOBILEGL_MAGMA_R11G11B10F_FALLBACK' "$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_COHERENT_AS_FLUSH' "$plugin_resource_text" 'V2 coherent-as-flush toggle'
+require 'key:MOBILEGL_USE_ANGLE' "$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
+  exit 1
+fi
+
+if ! grep -Eq '^lib/[^/]+/libMobileGL\.so$' <<<"$plugin_contents"; then
+  echo '::error::Plugin APK does not contain libMobileGL.so' >&2
+  exit 1
+fi
+
+require 'top.mobilegl.plugin.trace' "$trace_manifest" 'trace package name'
+require 'top.mobilegl.plugin.TRACE_REPLAY' "$trace_manifest" 'trace replay action'
+if grep -Fq 'fclPlugin' <<<"$trace_manifest"; then
+  echo '::error::Trace APK must not advertise renderer-plugin metadata' >&2
+  exit 1
+fi
+if grep -Fq 'android.intent.action.MAIN' <<<"$trace_manifest"; then
+  echo '::error::Trace APK must not expose a launcher activity' >&2
+  exit 1
+fi
+
+echo 'Validated unified MobileGL plugin APK and isolated trace APK.'

.github/workflows/apk.yml

@@ -0,0 +1,560 @@
+name: MobileGL APK
+
+on:
+  push:
+    branches:
+      - dev
+      - Feat/Backend-Direct-GLES
+      - Feat/Backend-Direct-Vulkan
+  workflow_dispatch:
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    env:
+      CCACHE_BASEDIR: ${{ github.workspace }}
+      CCACHE_COMPRESS: "true"
+      CCACHE_DIR: ${{ github.workspace }}/.ccache
+      CCACHE_MAXSIZE: 4G
+      CCACHE_NOHASHDIR: "true"
+      MOBILEGL_CMAKE_COMPILER_LAUNCHER: ccache
+
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+        with:
+          submodules: recursive
+
+      - name: Set artifact metadata
+        run: |
+          echo "date_today=$(date +'%Y-%m-%d')" >> "$GITHUB_ENV"
+
+      - name: Set up JDK
+        uses: actions/setup-java@v5
+        with:
+          distribution: zulu
+          java-version: '17'
+
+      - name: Setup Gradle
+        uses: gradle/actions/setup-gradle@v6
+        with:
+          gradle-version: 8.10.2
+
+      - name: Restore ccache
+        uses: actions/cache@v5
+        with:
+          path: .ccache
+          key: ${{ runner.os }}-apk-${{ github.job }}-ccache-${{ github.ref_name }}-${{ github.run_id }}
+          restore-keys: |
+            ${{ runner.os }}-apk-${{ github.job }}-ccache-${{ github.ref_name }}-
+            ${{ runner.os }}-apk-${{ github.job }}-ccache-
+
+      - name: Install ccache
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y ccache
+          ccache --version
+
+      - name: Setup Android SDK
+        uses: android-actions/setup-android@v4
+        with:
+          accept-android-sdk-licenses: false
+
+      - name: Accept Android SDK licenses
+        run: yes | sdkmanager --licenses >/dev/null
+
+      - name: Install Android NDK
+        run: |
+          sdkmanager "ndk;27.3.13750724"
+          echo "ndk.dir=$ANDROID_HOME/ndk/27.3.13750724" >> android-plugin/local.properties
+
+      - name: Update glslang external sources
+        working-directory: 3rdparty/glslang
+        run: python update_glslang_sources.py
+
+      - name: Build plugin APK
+        run: gradle --no-daemon -p android-plugin :app:assemblePluginRelease -Pmobilegl.apkSuffix="${GITHUB_SHA}" -Pmobilegl.logLevel=MOBILEGL_LOG_LEVEL_INFO --parallel --max-workers "$(nproc)"
+        env:
+          SIGNING_STORE_PASSWORD: ${{ secrets.SIGNING_STORE_PASSWORD }}
+          SIGNING_KEY_ALIAS: ${{ secrets.SIGNING_KEY_ALIAS }}
+          SIGNING_KEY_PASSWORD: ${{ secrets.SIGNING_KEY_PASSWORD }}
+
+      - name: Download ANGLE x86_64 libraries
+        run: |
+          angle_dir="android-plugin/app/src/trace/jniLibs/x86_64"
+          rm -rf "${angle_dir}"
+          mkdir -p "${angle_dir}"
+
+          package_angle_variant() {
+            variant="$1"
+            commit="$2"
+            egl_sha="$3"
+            gles_sha="$4"
+            source_dir="${RUNNER_TEMP}/mobilegl-angle-${variant}"
+            base="https://raw.githubusercontent.com/FCL-Team/FoldCraftLauncher/${commit}/FCLauncher/src/main/jniLibs/x86_64"
+            mkdir -p "${source_dir}"
+            curl -L --fail --retry 3 -o "${source_dir}/libEGL_angle.so" "${base}/libEGL_angle.so"
+            curl -L --fail --retry 3 -o "${source_dir}/libGLESv2_angle.so" "${base}/libGLESv2_angle.so"
+            echo "${egl_sha}  ${source_dir}/libEGL_angle.so" | sha256sum -c -
+            echo "${gles_sha}  ${source_dir}/libGLESv2_angle.so" | sha256sum -c -
+            for library in libEGL_angle libGLESv2_angle; do
+              filename="${library}_${variant}.so"
+              cp "${source_dir}/${library}.so" "${angle_dir}/${filename}"
+            done
+          }
+
+          package_angle_variant \
+            ec889e6ea831 \
+            f2a3d510dffd8f6540a52e1a7d0c5787d151075b \
+            c41828768d089899fa058ec0bee711a91be88347f29bdb935223da6be1149c40 \
+            e4f820d99f94365c66df868c7740fef142fe5c0cd7c941790a9e30638857ca4d
+          package_angle_variant \
+            90a62123d794 \
+            bdcc96ac11c79001018ae4375eb73cb54a9f682f \
+            d0f4298ccc770cc801fc52e21733521646161e8a4adb3bd0052d9a1b57ee0ca8 \
+            66fdc867e552192d553d59095ea2e3cef4829de65c356f1fd826027b1905972e
+
+      - name: Build retrace APK
+        run: gradle --no-daemon -p android-plugin :app:assembleTraceRelease -Pmobilegl.apkSuffix="${GITHUB_SHA}" -Pmobilegl.abis=all -Pmobilegl.debuggableRelease=true -Pmobilegl.logLevel=MOBILEGL_LOG_LEVEL_INFO --parallel --max-workers "$(nproc)"
+        env:
+          SIGNING_STORE_PASSWORD: ${{ secrets.SIGNING_STORE_PASSWORD }}
+          SIGNING_KEY_ALIAS: ${{ secrets.SIGNING_KEY_ALIAS }}
+          SIGNING_KEY_PASSWORD: ${{ secrets.SIGNING_KEY_PASSWORD }}
+
+      - name: Show ccache stats
+        if: always()
+        run: ccache --show-stats
+
+      - name: Verify APK metadata and packaging
+        run: |
+          AAPT2="$(find "$ANDROID_HOME/build-tools" -name aapt2 -type f | sort -V | tail -n 1)"
+          plugin_apk="android-plugin/app/build/outputs/apk/plugin/release/MobileGL-plugin-release-${GITHUB_SHA}.apk"
+          trace_apk="android-plugin/app/build/outputs/apk/trace/release/MobileGL-plugin-trace-release-${GITHUB_SHA}.apk"
+          test -f "${plugin_apk}"
+          test -f "${trace_apk}"
+          bash .github/scripts/validate-plugin-apks.sh "$AAPT2" "$plugin_apk" "$trace_apk"
+
+      - name: Verify signed APKs
+        run: |
+          APKSIGNER="$(find "$ANDROID_HOME/build-tools" -name apksigner -type f | sort -V | tail -n 1)"
+          mapfile -t APKS < <(printf '%s\n' \
+            "android-plugin/app/build/outputs/apk/plugin/release/MobileGL-plugin-release-${GITHUB_SHA}.apk" \
+            "android-plugin/app/build/outputs/apk/trace/release/MobileGL-plugin-trace-release-${GITHUB_SHA}.apk")
+          for APK in "${APKS[@]}"; do
+            if [[ ! -f "$APK" ]]; then
+              echo "::error::Expected release APK was not produced: $APK"
+              exit 1
+            fi
+          done
+
+          for APK in "${APKS[@]}"; do
+            if [[ "$APK" == *-unsigned.apk ]]; then
+              echo "::error::Unsigned release APK produced: $APK"
+              exit 1
+            fi
+            "$APKSIGNER" verify --verbose "$APK"
+          done
+
+      - name: Upload plugin APK
+        uses: actions/upload-artifact@v7
+        with:
+          name: MobileGL-plugin-${{ env.date_today }}-${{ github.sha }}
+          path: android-plugin/app/build/outputs/apk/plugin/release/MobileGL-plugin-release-${{ github.sha }}.apk
+          archive: false
+          if-no-files-found: error
+
+      - name: Upload retrace APK
+        uses: actions/upload-artifact@v7
+        with:
+          name: MobileGL-retrace-apk-${{ env.date_today }}-${{ github.sha }}
+          path: android-plugin/app/build/outputs/apk/trace/release/MobileGL-plugin-trace-release-${{ github.sha }}.apk
+          archive: false
+          if-no-files-found: error
+
+  trace-cases:
+    name: trace case matrix
+    runs-on: ubuntu-latest
+    needs: build
+    outputs:
+      android: ${{ steps.trace-cases.outputs.android }}
+      names: ${{ steps.trace-cases.outputs.names }}
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Load trace cases
+        id: trace-cases
+        run: |
+          echo "android=$(python3 tools/trace_replay/trace_cases.py --ci --format github-apk)" >> "$GITHUB_OUTPUT"
+          echo "names=$(python3 tools/trace_replay/trace_cases.py --ci --format names)" >> "$GITHUB_OUTPUT"
+
+  trace-fixtures:
+    name: trace fixture (${{ matrix.case }})
+    runs-on: ubuntu-latest
+    needs: trace-cases
+    strategy:
+      fail-fast: false
+      max-parallel: 4
+      matrix:
+        case: ${{ fromJSON(needs.trace-cases.outputs.names) }}
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Fetch trace fixture
+        run: bash .github/scripts/fetch-trace-fixture-lfs.sh '${{ matrix.case }}'
+
+      - name: Stage trace fixture
+        run: |
+          safe_case="$(printf '%s' '${{ matrix.case }}' | sed 's/[^A-Za-z0-9._-]/_/g')"
+          stage_dir="trace-fixtures/${safe_case}"
+          mkdir -p "${stage_dir}"
+          python3 tools/trace_replay/trace_cases.py --format fixture-files --case '${{ matrix.case }}' |
+            while IFS= read -r file; do
+              cp "${file}" "${stage_dir}/"
+            done
+
+      - name: Upload trace fixture
+        uses: actions/upload-artifact@v7
+        with:
+          name: MobileGL-trace-fixture-${{ matrix.case }}
+          path: trace-fixtures/**
+          if-no-files-found: error
+
+  android-avd:
+    name: android avd image
+    runs-on: ubuntu-latest
+    env:
+      AVD_NAME: mobilegl-ci
+      ANDROID_AVD_HOME: ${{ github.workspace }}/.android/avd
+      ANDROID_HOME: ${{ github.workspace }}/.android/sdk
+      ANDROID_SDK_ROOT: ${{ github.workspace }}/.android/sdk
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Setup Android SDK
+        uses: android-actions/setup-android@v4
+        with:
+          accept-android-sdk-licenses: false
+
+      - name: Accept Android SDK licenses
+        run: yes | sdkmanager --licenses >/dev/null
+
+      - name: Restore Android AVD cache
+        id: android-avd-cache
+        uses: actions/cache@v5
+        with:
+          path: |
+            ${{ env.ANDROID_AVD_HOME }}
+            ${{ env.ANDROID_SDK_ROOT }}/emulator
+            ${{ env.ANDROID_SDK_ROOT }}/platform-tools
+            ${{ env.ANDROID_SDK_ROOT }}/platforms/android-35
+            ${{ env.ANDROID_SDK_ROOT }}/system-images/android-35/google_apis/x86_64
+          key: ${{ runner.os }}-mobilegl-avd-api35-google_apis-x86_64-pixel_6-v2-${{ hashFiles('android-plugin/run-avd-ci.sh') }}
+
+      - name: Create AVD
+        if: steps.android-avd-cache.outputs.cache-hit != 'true'
+        run: |
+          sh android-plugin/run-avd-ci.sh create \
+            --api-level 35 \
+            --target google_apis \
+            --arch x86_64 \
+            --profile pixel_6 \
+            --avd-name "${AVD_NAME}"
+
+  retrace:
+    name: retrace (${{ matrix.backend.name }}, ${{ matrix.case.name }})
+    runs-on: ubuntu-latest
+    needs:
+      - build
+      - android-avd
+      - trace-cases
+      - trace-fixtures
+    if: ${{ always() && needs.build.result == 'success' && needs.android-avd.result == 'success' && needs.trace-cases.result == 'success' }}
+    timeout-minutes: 75
+    env:
+      AVD_NAME: mobilegl-ci
+      ANDROID_AVD_HOME: ${{ github.workspace }}/.android/avd
+      ANDROID_HOME: ${{ github.workspace }}/.android/sdk
+      ANDROID_SDK_ROOT: ${{ github.workspace }}/.android/sdk
+    strategy:
+      fail-fast: false
+      max-parallel: 4
+      matrix:
+        backend:
+          - name: DirectGLES
+            gpu: software
+          - name: DirectVulkan
+            gpu: lavapipe
+        case: ${{ fromJSON(needs.trace-cases.outputs.android) }}
+    steps:
+      - name: Set Swap Space
+        uses: pierotofy/set-swap-space@v1.0
+        with:
+          swap-size-gb: 8
+
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Download trace fixture
+        uses: actions/download-artifact@v8
+        with:
+          name: MobileGL-trace-fixture-${{ matrix.case.name }}
+          path: trace-fixture-download
+
+      - name: Install trace fixture
+        run: |
+          mkdir -p tools/trace_replay/fixtures
+          find trace-fixture-download -type f -exec cp {} tools/trace_replay/fixtures/ \;
+
+      - name: Set artifact metadata
+        run: |
+          echo "date_today=$(date +'%Y-%m-%d')" >> "$GITHUB_ENV"
+          echo "EMULATOR_LOG=${RUNNER_TEMP}/mobilegl-emulator.log" >> "$GITHUB_ENV"
+          echo "EMULATOR_PID_FILE=${RUNNER_TEMP}/mobilegl-emulator.pid" >> "$GITHUB_ENV"
+
+      - name: Setup Android SDK
+        uses: android-actions/setup-android@v4
+        with:
+          accept-android-sdk-licenses: false
+
+      - name: Accept Android SDK licenses
+        run: yes | sdkmanager --licenses >/dev/null
+
+      - name: Restore Android AVD cache
+        id: android-avd-cache
+        uses: actions/cache/restore@v5
+        with:
+          path: |
+            ${{ env.ANDROID_AVD_HOME }}
+            ${{ env.ANDROID_SDK_ROOT }}/emulator
+            ${{ env.ANDROID_SDK_ROOT }}/platform-tools
+            ${{ env.ANDROID_SDK_ROOT }}/platforms/android-35
+            ${{ env.ANDROID_SDK_ROOT }}/system-images/android-35/google_apis/x86_64
+          key: ${{ runner.os }}-mobilegl-avd-api35-google_apis-x86_64-pixel_6-v2-${{ hashFiles('android-plugin/run-avd-ci.sh') }}
+
+      - name: Download retrace APK
+        uses: actions/download-artifact@v8
+        with:
+          name: MobileGL-plugin-trace-release-${{ github.sha }}.apk
+          path: android-retrace-apks
+
+      - name: Enable KVM
+        run: |
+          echo 'KERNEL=="kvm", GROUP="kvm", MODE="0666", OPTIONS+="static_node=kvm"' | sudo tee /etc/udev/rules.d/99-kvm4all.rules
+          sudo udevadm control --reload-rules
+          sudo udevadm trigger --name-match=kvm
+
+      - name: Create AVD
+        if: steps.android-avd-cache.outputs.cache-hit != 'true'
+        run: |
+          sh android-plugin/run-avd-ci.sh create \
+            --api-level 35 \
+            --target google_apis \
+            --arch x86_64 \
+            --profile pixel_6 \
+            --avd-name "${AVD_NAME}"
+
+      - name: Launch Emulator
+        run: |
+          sh android-plugin/run-avd-ci.sh start \
+            --avd-name "${AVD_NAME}" \
+            --gpu "${{ matrix.backend.gpu }}" \
+            --emulator-log "${EMULATOR_LOG}" \
+            --pid-file "${EMULATOR_PID_FILE}" \
+            --boot-timeout 300
+
+      - name: Retrace and validate
+        env:
+          MOBILEGL_USE_ANGLE: ${{ matrix.backend.name == 'DirectGLES' && '1' || '0' }}
+          MOBILEGL_TRACE_ANGLE_VARIANT: ${{ matrix.case.name == 'minecraft-1.21.4-fabric-iris-bliss-in-world' && '90a62123d794' || 'ec889e6ea831' }}
+          MOBILEGL_MAGMA_R11G11B10F_FALLBACK: ${{ matrix.backend.name == 'DirectVulkan' && '1' || '0' }}
+        run: |
+          apk_file="android-retrace-apks/MobileGL-plugin-trace-release-${GITHUB_SHA}.apk"
+          test -f "${apk_file}"
+          extra_retrace_args=()
+          # Bliss needs the newer signed ANGLE variant plus sampler mipmap
+          # min-filter downgrading on ANGLE llvmpipe.
+          if [ "${{ 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.case.coherent_as_flush || false }}" = "true" ]; then
+            extra_retrace_args+=(--coherent-as-flush)
+          fi
+
+          run_retrace() {
+            timeout "$(( ${{ matrix.case.timeout_seconds }} + 300 ))" sh android-plugin/trace-replay-ci.sh \
+              --apk-file "${apk_file}" \
+              --package top.mobilegl.plugin.trace \
+              --backend "${{ matrix.backend.name }}" \
+              --result-root android-retrace-result \
+              --fixture-root android-retrace-fixture \
+              --case "${{ matrix.case.name }}" \
+              --trace-archive "${{ matrix.case.trace_archive }}" \
+              --trace-file "${{ matrix.case.trace_file }}" \
+              --golden "${{ matrix.case.golden }}" \
+              --alternate-golden "${{ matrix.case.alternate_golden || '' }}" \
+              --target-call "${{ matrix.case.target_call }}" \
+              --width "${{ matrix.case.width }}" \
+              --height "${{ matrix.case.height }}" \
+              --ssim-threshold "${{ matrix.case.ssim_threshold || '0.99' }}" \
+              --crop-x "${{ matrix.case.crop_x }}" \
+              --crop-y "${{ matrix.case.crop_y }}" \
+              --crop-width "${{ matrix.case.crop_width }}" \
+              --crop-height "${{ matrix.case.crop_height }}" \
+              --timeout-seconds "${{ matrix.case.timeout_seconds }}" \
+              "${extra_retrace_args[@]}"
+          }
+
+          retrace_status=0
+          run_retrace || retrace_status=$?
+          if [ "${retrace_status}" -eq 75 ]; then
+            echo "::warning::Android emulator infrastructure failed; restarting it and retrying this retrace once."
+            sh android-plugin/run-avd-ci.sh stop \
+              --avd-name "${AVD_NAME}" \
+              --emulator-log "${EMULATOR_LOG}" \
+              --pid-file "${EMULATOR_PID_FILE}"
+            adb kill-server || true
+            sleep 2
+            sh android-plugin/run-avd-ci.sh start \
+              --avd-name "${AVD_NAME}" \
+              --gpu "${{ matrix.backend.gpu }}" \
+              --emulator-log "${EMULATOR_LOG}" \
+              --pid-file "${EMULATOR_PID_FILE}" \
+              --boot-timeout 300
+            run_retrace
+          elif [ "${retrace_status}" -ne 0 ]; then
+            exit "${retrace_status}"
+          fi
+
+      - name: Collect retrace summary inputs
+        if: always()
+        run: |
+          safe_case="$(printf '%s' '${{ matrix.case.name }}' | sed 's/[^A-Za-z0-9._-]/_/g')"
+          result_dir="android-retrace-result/${safe_case}-${{ matrix.backend.name }}"
+          mkdir -p "${result_dir}"
+          if [ -s "${{ matrix.case.golden }}" ]; then
+            cp "${{ matrix.case.golden }}" "${result_dir}/${safe_case}-${{ matrix.backend.name }}-golden.png"
+          fi
+          if [ -n "${{ matrix.case.alternate_golden || '' }}" ] && [ -s "${{ matrix.case.alternate_golden || '' }}" ]; then
+            cp "${{ matrix.case.alternate_golden || '' }}" "${result_dir}/${safe_case}-${{ matrix.backend.name }}-alternate-golden.png"
+          fi
+
+      - name: Collect emulator diagnostics
+        if: always()
+        run: |
+          mkdir -p android-retrace-result/diagnostics
+          adb devices -l > android-retrace-result/diagnostics/adb-devices.txt || true
+          timeout 30 adb logcat -d -t 1000 > android-retrace-result/diagnostics/logcat.txt || true
+          if [ -f "${EMULATOR_LOG}" ]; then
+            cp "${EMULATOR_LOG}" android-retrace-result/diagnostics/emulator.log
+          fi
+
+      - name: Stop Emulator
+        if: always()
+        run: |
+          sh android-plugin/run-avd-ci.sh stop \
+            --avd-name "${AVD_NAME}" \
+            --emulator-log "${EMULATOR_LOG}" \
+            --pid-file "${EMULATOR_PID_FILE}"
+
+      - name: Upload Android retrace result
+        if: always()
+        uses: actions/upload-artifact@v7
+        with:
+          name: MobileGL-android-retrace-result-${{ env.date_today }}-${{ github.sha }}-${{ matrix.backend.name }}-${{ matrix.case.name }}
+          path: android-retrace-result/**
+          if-no-files-found: warn
+
+  retrace-summary:
+    name: retrace summary
+    runs-on: ubuntu-latest
+    needs: retrace
+    if: always()
+
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Set artifact metadata
+        run: |
+          echo "date_today=$(date +'%Y-%m-%d')" >> "$GITHUB_ENV"
+
+      - name: Set up Node.js
+        uses: actions/setup-node@v7
+        with:
+          node-version: '22'
+
+      - name: Download Android retrace results
+        uses: actions/download-artifact@v8
+        with:
+          pattern: MobileGL-android-retrace-result-*
+          path: retrace-artifacts
+
+      - name: Render retrace summary
+        run: |
+          node tools/trace_replay/render_retrace_summary.mjs \
+            --input retrace-artifacts \
+            --output-dir android-retrace-summary \
+            --title "MobileGL Android retrace overview" \
+            --group-label "Android Emulator" \
+            --html mobilegl-android-retrace-overview.html
+
+      - name: Upload Android retrace summary
+        uses: actions/upload-artifact@v7
+        with:
+          path: android-retrace-summary/mobilegl-android-retrace-overview.html
+          archive: false
+          if-no-files-found: error
+
+  remove-artifact-clutter:
+    name: remove artifact clutter
+    runs-on: ubuntu-latest
+    needs: retrace-summary
+    if: always()
+    permissions:
+      actions: write
+    steps:
+      - name: Delete intermediate Android retrace artifacts
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          declare -A failed_cases=()
+          while IFS= read -r job_name; do
+            case_name="${job_name#retrace (*, }"
+            case_name="${case_name%)}"
+            failed_cases["${case_name}"]=1
+          done < <(
+            gh api --paginate "repos/${GITHUB_REPOSITORY}/actions/runs/${GITHUB_RUN_ID}/jobs?per_page=100" \
+              --jq '.jobs[] | select(.name | startswith("retrace (")) | select(.conclusion == "failure" or .conclusion == "cancelled" or .conclusion == "timed_out" or .conclusion == "action_required") | .name'
+          )
+
+          if ((${#failed_cases[@]})); then
+            echo "Retaining fixtures for failed retrace case(s):"
+            printf '  %s\n' "${!failed_cases[@]}"
+          else
+            echo "All retrace jobs succeeded; no fixtures need to be retained."
+          fi
+
+          deleted=0
+          retained=0
+          while IFS=$'\t' read -r artifact_id artifact_name; do
+            if [[ "${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
+              fi
+            fi
+
+            echo "Deleting ${artifact_name} (${artifact_id})"
+            gh api --method DELETE "repos/${GITHUB_REPOSITORY}/actions/artifacts/${artifact_id}"
+            ((deleted += 1))
+          done < <(
+            gh api --paginate "repos/${GITHUB_REPOSITORY}/actions/runs/${GITHUB_RUN_ID}/artifacts?per_page=100" \
+              --jq '.artifacts[] | select(.name | startswith("MobileGL-trace-fixture-") or startswith("MobileGL-android-retrace-result-") or startswith("trace-fixture-") or startswith("retrace-result-")) | [.id, .name] | @tsv'
+          )
+
+          echo "Deleted ${deleted} intermediate Android artifact(s); retained ${retained} failed-retrace fixture(s)."

.github/workflows/benchmark.yml

@@ -1,64 +0,0 @@
-name: Benchmark
-
-on:
-  push:
-    branches: 
-      - dev
-      - Feat/Backend-Direct-GLES
-      - Feat/Backend-Direct-Vulkan
-
-jobs:
-  benchmark:
-    runs-on: ubuntu-latest
-    env:
-      # BENCH_ROOT: ${{github.workspace}}/MobileGL/MG_Benchmark
-      BENCH_ROOT: ${{github.workspace}}
-
-    steps:
-      - name: Set Swap Space
-        uses: pierotofy/set-swap-space@master
-        with:
-          swap-size-gb: 32
-
-      - name: Checkout repo
-        uses: actions/checkout@v4
-        with:
-          submodules: true
-
-      - name: Get CMake
-        uses: lukka/get-cmake@latest
-
-      - name: Prepare Vulkan SDK
-        uses: humbletim/setup-vulkan-sdk@v1.2.1
-        with:
-          vulkan-query-version: 1.4.304.1
-          vulkan-components: Vulkan-Headers, Vulkan-Loader
-          vulkan-use-cache: true
-
-      - name: Update glslang external sources
-        working-directory: ${{env.BENCH_ROOT}}/3rdparty/glslang
-        run: python update_glslang_sources.py
-
-      - name: Install clang-20
-        run: |
-          sudo apt-get update
-          sudo apt-get install -y clang-20 clang++-20 lld-20 libc++-20-dev libc++abi-20-dev libvulkan-dev
-
-      - name: Show installed toolchain
-        run: |
-          clang-20 --version
-          clang++-20 --version
-          ld.lld-20 --version || ld.lld --version || true
-          dpkg -l 'libc++*' || true
-
-      - name: Configure CMake
-        working-directory: ${{env.BENCH_ROOT}}
-        run: cmake -S . -B build-bench -G Ninja -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang-20 -DCMAKE_CXX_COMPILER=clang++-20 -DBENCHMARK_DOWNLOAD_DEPENDENCIES=ON -DBENCHMARK_ENABLE_TESTING=OFF -DMOBILEGL_BUILD_TEST=OFF -DMOBILEGL_BUILD_BENCHMARK=ON -DCMAKE_POLICY_VERSION_MINIMUM=3.5
-
-      - name: Build
-        working-directory: ${{env.BENCH_ROOT}}/build-bench
-        run: cmake --build .
-
-      - name: Benchmark
-        working-directory: ${{env.BENCH_ROOT}}/build-bench/MobileGL/MG_Benchmark
-        run: ctest -V -C Release

.github/workflows/test.yml

@@ -6,27 +6,41 @@ on:
       - dev
       - Feat/Backend-Direct-GLES
       - Feat/Backend-Direct-Vulkan
+  workflow_dispatch:
 
 jobs:
-  test:
+  build-linux:
     runs-on: ubuntu-latest
     env:
-      # TEST_ROOT: ${{github.workspace}}/MobileGL/MG_Test
-      TEST_ROOT: ${{github.workspace}}
+      BUILD_DIR: build-linux
+      CCACHE_BASEDIR: ${{ github.workspace }}
+      CCACHE_COMPRESS: "true"
+      CCACHE_DIR: ${{ github.workspace }}/.ccache
+      CCACHE_MAXSIZE: 4G
+      CCACHE_NOHASHDIR: "true"
 
     steps:
       - name: Set Swap Space
-        uses: pierotofy/set-swap-space@master
+        uses: pierotofy/set-swap-space@v1.0
         with:
           swap-size-gb: 32
 
       - name: Checkout repo
-        uses: actions/checkout@v4
+        uses: actions/checkout@v6
         with:
-          submodules: true
+          submodules: recursive
 
       - name: Get CMake
-        uses: lukka/get-cmake@latest
+        uses: lukka/get-cmake@v4.3.3
+
+      - name: Restore ccache
+        uses: actions/cache@v5
+        with:
+          path: .ccache
+          key: ${{ runner.os }}-test-${{ github.job }}-ccache-${{ github.ref_name }}-${{ github.run_id }}
+          restore-keys: |
+            ${{ runner.os }}-test-${{ github.job }}-ccache-${{ github.ref_name }}-
+            ${{ runner.os }}-test-${{ github.job }}-ccache-
 
       - name: Prepare Vulkan SDK
         uses: humbletim/setup-vulkan-sdk@v1.2.1
@@ -36,39 +50,520 @@ jobs:
           vulkan-use-cache: true
 
       - name: Update glslang external sources
-        working-directory: ${{env.TEST_ROOT}}/3rdparty/glslang
+        working-directory: 3rdparty/glslang
         run: python update_glslang_sources.py
 
-      - name: Install clang-20
+      - name: Install build dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install -y clang-20 clang++-20 lld-20 libc++-20-dev libc++abi-20-dev libvulkan-dev
+          sudo apt-get install -y ccache clang-20 clang++-20 lld-20 libc++-20-dev libc++abi-20-dev libvulkan-dev libegl1-mesa-dev libgles2-mesa-dev libgl1-mesa-dri mesa-vulkan-drivers ninja-build
 
       - name: Show installed toolchain
         run: |
+          ccache --version
           clang-20 --version
           clang++-20 --version
           ld.lld-20 --version || ld.lld --version || true
-          dpkg -l 'libc++*' || true
-          
+          dpkg -l 'libc++*' 'libegl*' 'libgles*' 'mesa*' 'vulkan*' || true
+
       - name: Configure CMake
-        working-directory: ${{env.TEST_ROOT}}
         run: |
-          if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" == "true" ]; then
-            cmake -S . -B build-test -G Ninja -DCMAKE_C_COMPILER=clang-20 -DCMAKE_CXX_COMPILER=clang++-20 -DCMAKE_BUILD_TYPE=Debug -DMOBILEGL_BUILD_TEST=ON -DMOBILEGL_BUILD_BENCHMARK=OFF -DCMAKE_POLICY_VERSION_MINIMUM=3.5
+          if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" = "true" ]; then
+            BUILD_TYPE=Debug
           else
-            cmake -S . -B build-test -G Ninja -DCMAKE_C_COMPILER=clang-20 -DCMAKE_CXX_COMPILER=clang++-20 -DCMAKE_BUILD_TYPE=Release -DMOBILEGL_BUILD_TEST=ON -DMOBILEGL_BUILD_BENCHMARK=OFF -DCMAKE_POLICY_VERSION_MINIMUM=3.5
+            BUILD_TYPE=Release
           fi
-          
+
+          cmake -S . -B "${BUILD_DIR}" -G Ninja \
+            -DCMAKE_C_COMPILER=clang-20 \
+            -DCMAKE_CXX_COMPILER=clang++-20 \
+            -DCMAKE_C_COMPILER_LAUNCHER=ccache \
+            -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+            -DCMAKE_BUILD_TYPE="${BUILD_TYPE}" \
+            -DMOBILEGL_LOG_ACTIVE_LEVEL=MOBILEGL_LOG_LEVEL_INFO \
+            -DMOBILEGL_BUILD_TEST=ON \
+            -DMOBILEGL_BUILD_BENCHMARK=ON \
+            -DMOBILEGL_BUILD_TRACE_REPLAY=OFF \
+            -DBENCHMARK_DOWNLOAD_DEPENDENCIES=ON \
+            -DBENCHMARK_ENABLE_TESTING=OFF \
+            -DCMAKE_POLICY_VERSION_MINIMUM=3.5
+
       - name: Build
-        working-directory: ${{env.TEST_ROOT}}/build-test
-        run: cmake --build .
+        run: cmake --build "${BUILD_DIR}" --parallel "$(nproc)"
+
+      - name: Show ccache stats
+        if: always()
+        run: ccache --show-stats
+
+      - name: Package Linux runtime
+        run: |
+          mkdir -p ci-artifacts
+          mapfile -t SHARED_LIBS < <(find "${BUILD_DIR}" -type f \( -name '*.so' -o -name '*.so.*' \) -print | sort)
+          tar \
+            --exclude='*/CMakeFiles' \
+            --exclude='*.o' \
+            --exclude='*.a' \
+            --exclude='*.ninja*' \
+            --exclude='build.ninja' \
+            --exclude='cmake_install.cmake' \
+            -czf ci-artifacts/mobilegl-linux-runtime.tgz \
+            "${BUILD_DIR}/CTestTestfile.cmake" \
+            "${BUILD_DIR}/MobileGL/MG_Test" \
+            "${BUILD_DIR}/MobileGL/MG_Benchmark" \
+            "${SHARED_LIBS[@]}"
+
+      - name: Upload Linux runtime
+        uses: actions/upload-artifact@v7
+        with:
+          name: mobilegl-linux-runtime
+          path: ci-artifacts/mobilegl-linux-runtime.tgz
+          if-no-files-found: error
+
+  test:
+    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
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libvulkan1 libegl1 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: Test
-        working-directory: ${{env.TEST_ROOT}}/build-test/MobileGL/MG_Test
+        working-directory: build-linux
         run: |
-          if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" == "true" ]; then
-            ctest -V
+          if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" = "true" ]; then
+            ctest -V -L unit --no-tests=error
           else
-            ctest
+            ctest --output-on-failure -L unit --no-tests=error
           fi
+
+  benchmark:
+    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
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libvulkan1 libegl1 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: Benchmark
+        working-directory: build-linux
+        run: ctest -V -C Release -L benchmark --no-tests=error
+
+  build-retrace:
+    runs-on: ubuntu-latest
+    needs:
+      - build-linux
+      - test
+      - benchmark
+    env:
+      BUILD_DIR: build-retrace
+      CCACHE_BASEDIR: ${{ github.workspace }}
+      CCACHE_COMPRESS: "true"
+      CCACHE_DIR: ${{ github.workspace }}/.ccache
+      CCACHE_MAXSIZE: 4G
+      CCACHE_NOHASHDIR: "true"
+      MOBILEGL_LIBRARY: ${{ github.workspace }}/build-linux/libMobileGL.so
+
+    steps:
+      - name: Set Swap Space
+        uses: pierotofy/set-swap-space@v1.0
+        with:
+          swap-size-gb: 32
+
+      - name: Checkout repo
+        uses: actions/checkout@v6
+        with:
+          submodules: recursive
+
+      - name: Get CMake
+        uses: lukka/get-cmake@v4.3.3
+
+      - name: Restore ccache
+        uses: actions/cache@v5
+        with:
+          path: .ccache
+          key: ${{ runner.os }}-test-${{ github.job }}-ccache-${{ github.ref_name }}-${{ github.run_id }}
+          restore-keys: |
+            ${{ runner.os }}-test-${{ github.job }}-ccache-${{ github.ref_name }}-
+            ${{ runner.os }}-test-${{ github.job }}-ccache-
+
+      - name: Prepare Vulkan SDK
+        uses: humbletim/setup-vulkan-sdk@v1.2.1
+        with:
+          vulkan-query-version: 1.4.304.1
+          vulkan-components: Vulkan-Headers, Vulkan-Loader
+          vulkan-use-cache: true
+
+      - name: Update glslang external sources
+        working-directory: 3rdparty/glslang
+        run: python update_glslang_sources.py
+
+      - name: Install dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y ccache clang-20 clang++-20 lld-20 libc++-20-dev libc++abi-20-dev libvulkan-dev libegl1-mesa-dev libgles2-mesa-dev libgl1-mesa-dri mesa-vulkan-drivers ninja-build
+
+      - name: Show installed toolchain
+        run: |
+          ccache --version
+          clang-20 --version
+          clang++-20 --version
+          ld.lld-20 --version || ld.lld --version || true
+          dpkg -l 'libc++*' 'libegl*' 'libgles*' 'mesa*' 'vulkan*' || true
+
+      - 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
+          test -f "${MOBILEGL_LIBRARY}"
+
+      - name: Configure CMake
+        run: |
+          if [ "${{ secrets.ACTIONS_STEP_DEBUG }}" = "true" ]; then
+            BUILD_TYPE=Debug
+          else
+            BUILD_TYPE=Release
+          fi
+
+          cmake -S . -B "${BUILD_DIR}" -G Ninja \
+            -DCMAKE_C_COMPILER=clang-20 \
+            -DCMAKE_CXX_COMPILER=clang++-20 \
+            -DCMAKE_C_COMPILER_LAUNCHER=ccache \
+            -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+            -DCMAKE_BUILD_TYPE="${BUILD_TYPE}" \
+            -DMOBILEGL_LOG_ACTIVE_LEVEL=MOBILEGL_LOG_LEVEL_INFO \
+            -DMOBILEGL_BUILD_TEST=OFF \
+            -DMOBILEGL_BUILD_BENCHMARK=OFF \
+            -DMOBILEGL_BUILD_TRACE_REPLAY=ON \
+            -DMOBILEGL_TRACE_REPLAY_MOBILEGL_LIBRARY="${MOBILEGL_LIBRARY}" \
+            -DCMAKE_POLICY_VERSION_MINIMUM=3.5
+
+      - name: Build trace replay
+        run: cmake --build "${BUILD_DIR}" --target mobilegl_trace_replay --parallel "$(nproc)"
+
+      - name: Show ccache stats
+        if: always()
+        run: ccache --show-stats
+
+      - name: Normalize CTest command paths
+        run: |
+          python - <<'PY'
+          from pathlib import Path
+          import re
+
+          for path in Path('build-retrace').rglob('CTestTestfile.cmake'):
+              text = path.read_text()
+              text = re.sub(r'"[^"]*/cmake-[^"]*/bin/cmake"', '"cmake"', text)
+              path.write_text(text)
+          PY
+
+      - name: Package trace replay
+        run: |
+          mkdir -p ci-artifacts
+          tar -czf ci-artifacts/mobilegl-trace-replay.tgz \
+            build-retrace/tools/trace_replay/mobilegl_trace_replay \
+            build-retrace/tools/trace_replay/CTestTestfile.cmake
+
+      - name: Upload trace replay
+        uses: actions/upload-artifact@v7
+        with:
+          name: mobilegl-trace-replay
+          path: ci-artifacts/mobilegl-trace-replay.tgz
+          if-no-files-found: error
+
+  trace-cases:
+    name: trace case matrix
+    runs-on: ubuntu-latest
+    needs:
+      - test
+      - benchmark
+    outputs:
+      names: ${{ steps.trace-cases.outputs.names }}
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Load trace cases
+        id: trace-cases
+        run: echo "names=$(python3 tools/trace_replay/trace_cases.py --ci --format names)" >> "$GITHUB_OUTPUT"
+
+  trace-fixtures:
+    name: trace fixture (${{ matrix.case }})
+    runs-on: ubuntu-latest
+    needs: trace-cases
+    strategy:
+      fail-fast: false
+      max-parallel: 4
+      matrix:
+        case: ${{ fromJSON(needs.trace-cases.outputs.names) }}
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Fetch trace fixture
+        run: bash .github/scripts/fetch-trace-fixture-lfs.sh '${{ matrix.case }}'
+
+      - name: Stage trace fixture
+        run: |
+          safe_case="$(printf '%s' '${{ matrix.case }}' | sed 's/[^A-Za-z0-9._-]/_/g')"
+          stage_dir="trace-fixtures/${safe_case}"
+          mkdir -p "${stage_dir}"
+          python3 tools/trace_replay/trace_cases.py --format fixture-files --case '${{ matrix.case }}' |
+            while IFS= read -r file; do
+              cp "${file}" "${stage_dir}/"
+            done
+
+      - name: Upload trace fixture
+        uses: actions/upload-artifact@v7
+        with:
+          name: trace-fixture-${{ matrix.case }}
+          path: trace-fixtures/**
+          if-no-files-found: error
+
+  retrace:
+    name: retrace (${{ matrix.backend }}, ${{ matrix.case }})
+    runs-on: ubuntu-latest
+    needs:
+      - build-linux
+      - build-retrace
+      - trace-cases
+      - trace-fixtures
+    if: ${{ always() && needs.build-linux.result == 'success' && needs.build-retrace.result == 'success' && needs.trace-cases.result == 'success' }}
+    strategy:
+      fail-fast: false
+      max-parallel: 4
+      matrix:
+        backend:
+          - DirectGLES
+          - DirectVulkan
+        case: ${{ fromJSON(needs.trace-cases.outputs.names) }}
+
+    steps:
+      - name: Set Swap Space
+        uses: pierotofy/set-swap-space@v1.0
+        with:
+          swap-size-gb: 16
+
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Download trace fixture
+        uses: actions/download-artifact@v8
+        with:
+          name: trace-fixture-${{ matrix.case }}
+          path: trace-fixture-download
+
+      - name: Install trace fixture
+        run: |
+          mkdir -p tools/trace_replay/fixtures
+          find trace-fixture-download -type f -exec cp {} tools/trace_replay/fixtures/ \;
+
+      - name: Get CMake
+        uses: lukka/get-cmake@v4.3.3
+
+      - name: Install runtime dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libvulkan1 libegl1-mesa-dev libgles2-mesa-dev libgl1-mesa-dri mesa-vulkan-drivers
+          test -e /usr/lib/x86_64-linux-gnu/libEGL.so
+          test -e /usr/lib/x86_64-linux-gnu/libGLESv2.so
+
+      - name: Download Linux runtime
+        uses: actions/download-artifact@v8
+        with:
+          name: mobilegl-linux-runtime
+          path: .
+
+      - name: Download trace replay
+        uses: actions/download-artifact@v8
+        with:
+          name: mobilegl-trace-replay
+          path: .
+
+      - name: Unpack retrace runtime
+        run: |
+          tar -xzf mobilegl-linux-runtime.tgz
+          tar -xzf mobilegl-trace-replay.tgz
+          test -f build-linux/libMobileGL.so
+          test -f build-retrace/tools/trace_replay/mobilegl_trace_replay
+
+      - name: Retrace and validate
+        working-directory: build-retrace/tools/trace_replay
+        run: |
+          if [ '${{ matrix.backend }}' = 'DirectVulkan' ]; then
+            export MOBILEGL_MAGMA_R11G11B10F_FALLBACK=1
+          fi
+          # The blended depth-write quirk auto-enables only on Qualcomm, which no CI
+          # runner has, so force it on for the OIT case it exists to fix. ForceOn
+          # bypasses only the vendor gate, so this exercises the real strip on
+          # lavapipe. The Android AVD lane deliberately leaves it off, keeping the
+          # unstripped path covered for the same trace.
+          if [ '${{ 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 actual image
+        if: always()
+        uses: actions/upload-artifact@v7
+        with:
+          name: retrace-result-${{ matrix.backend }}-${{ matrix.case }}
+          path: |
+            build-retrace/tools/trace_replay/${{ matrix.case }}/actual-images/**
+            build-retrace/tools/trace_replay/${{ matrix.case }}/${{ matrix.backend }}/output/**
+          if-no-files-found: warn
+
+  retrace-summary:
+    name: retrace summary
+    runs-on: ubuntu-latest
+    needs: retrace
+    if: ${{ always() && needs.retrace.result != 'skipped' }}
+
+    steps:
+      - name: Checkout repo
+        uses: actions/checkout@v6
+
+      - name: Set artifact metadata
+        run: |
+          echo "date_today=$(date +'%Y-%m-%d')" >> "$GITHUB_ENV"
+
+      - name: Set up Node.js
+        uses: actions/setup-node@v7
+        with:
+          node-version: '22'
+
+      - name: Download retrace results
+        uses: actions/download-artifact@v8
+        with:
+          pattern: retrace-result-*
+          path: retrace-artifacts
+
+      - name: Render retrace summary
+        run: |
+          node tools/trace_replay/render_retrace_summary.mjs \
+            --input retrace-artifacts \
+            --output-dir retrace-summary \
+            --title "MobileGL Linux retrace overview" \
+            --group-label "Linux" \
+            --html mobilegl-linux-retrace-overview.html
+
+      - name: Upload retrace summary
+        uses: actions/upload-artifact@v7
+        with:
+          path: retrace-summary/mobilegl-linux-retrace-overview.html
+          archive: false
+          if-no-files-found: error
+
+  remove-artifact-clutter:
+    name: remove artifact clutter
+    runs-on: ubuntu-latest
+    needs: retrace-summary
+    if: always()
+    permissions:
+      actions: write
+    steps:
+      - name: Delete intermediate Linux retrace artifacts
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          declare -A failed_cases=()
+          while IFS= read -r job_name; do
+            case_name="${job_name#retrace (*, }"
+            case_name="${case_name%)}"
+            failed_cases["${case_name}"]=1
+          done < <(
+            gh api --paginate "repos/${GITHUB_REPOSITORY}/actions/runs/${GITHUB_RUN_ID}/jobs?per_page=100" \
+              --jq '.jobs[] | select(.name | startswith("retrace (")) | select(.conclusion == "failure" or .conclusion == "cancelled" or .conclusion == "timed_out" or .conclusion == "action_required") | .name'
+          )
+
+          if ((${#failed_cases[@]})); then
+            echo "Retaining fixtures for failed retrace case(s):"
+            printf '  %s\n' "${!failed_cases[@]}"
+          else
+            echo "All retrace jobs succeeded; no fixtures need to be retained."
+          fi
+
+          deleted=0
+          retained=0
+          while IFS=$'\t' read -r artifact_id artifact_name; do
+            if [[ "${artifact_name}" == trace-fixture-* ]]; then
+              case_name="${artifact_name#trace-fixture-}"
+              if [[ -v "failed_cases[${case_name}]" ]]; then
+                echo "Retaining ${artifact_name} (${artifact_id}) for failed retrace."
+                ((retained += 1))
+                continue
+              fi
+            fi
+
+            echo "Deleting ${artifact_name} (${artifact_id})"
+            gh api --method DELETE "repos/${GITHUB_REPOSITORY}/actions/artifacts/${artifact_id}"
+            ((deleted += 1))
+          done < <(
+            gh api --paginate "repos/${GITHUB_REPOSITORY}/actions/runs/${GITHUB_RUN_ID}/artifacts?per_page=100" \
+              --jq '.artifacts[] | select(.name | startswith("trace-fixture-") or startswith("retrace-result-")) | [.id, .name] | @tsv'
+          )
+
+          echo "Deleted ${deleted} intermediate Linux artifact(s); retained ${retained} failed-retrace fixture(s)."

.gitignore

@@ -18,4 +18,10 @@ MobileGL/MG_Test/build
 /cmake-build*
 .idea
 MobileGL/MG*/build*
-MobileGL/MG*/cmake-build*
+MobileGL/MG*/cmake-build*
+/android-plugin/.gradle
+/android-plugin/build
+/android-plugin/app/build
+/android-plugin/app/src/trace/jniLibs
+/android-plugin/local.properties
+tools/trace_replay/work/

.gitmodules

@@ -28,3 +28,6 @@
 [submodule "3rdparty/SPIRV-Reflect"]
 	path = 3rdparty/SPIRV-Reflect
 	url = https://github.com/KhronosGroup/SPIRV-Reflect.git
+[submodule "3rdparty/apitrace"]
+	path = 3rdparty/apitrace
+	url = https://github.com/MobileGL-Dev/apitrace.git

CMakeLists.txt

@@ -6,6 +6,11 @@ option(MOBILEGL_BUILD_TEST        "Build MobileGL tests"
 option(MOBILEGL_BUILD_BENCHMARK   "Build MobileGL benchmarks"                        ON )
 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)
+option(MOBILEGL_TRACE_ANGLE_VARIANTS "Enable signed trace-APK ANGLE variant loading" OFF)
+option(MOBILEGL_IOS               "Build MobileGL for iOS instead of macOS when APPLE is set" OFF)
+set(MOBILEGL_LOG_ACTIVE_LEVEL "MOBILEGL_LOG_LEVEL_INFO" CACHE STRING "MobileGL active log level macro")
+set(MOBILEGL_VULKAN_LIBRARY "" CACHE FILEPATH "Vulkan loader/MoltenVK library to link for iOS builds")
 
 if (ANDROID)
     set(MOBILEGL_BUILD_TEST       OFF CACHE BOOL "Build MobileGL tests"      FORCE)
@@ -179,11 +184,25 @@ set(SOURCE_FILES
     MobileGL/MG_Util/ShaderTranspiler/SpvcSession.cpp
     MobileGL/MG_Util/ShaderTranspiler/ShaderSourceProcessor.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/DecomposeWorkgroupVec3Pass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/DecoratePositionInvariantPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/LowerDrawParametersPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/RebaseInstanceIndexPass.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/FoldConstOffsetFor1DFetchPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/LowerClipDistanceForEsslPass.cpp
+    MobileGL/MG_Util/ShaderTranspiler/SpirvPasses/DefeatConstStructArrayLutPass.cpp
 
     MobileGL/MG_Util/BackendLoaders/OpenGL/Loader.cpp
     MobileGL/MG_Util/BackendLoaders/Vulkan/Loader.cpp
 
+    MobileGL/MG_Util/SelfTest/DriverPost.cpp
+
     MobileGL/MG_Util/Texture/PixelStoreProcessor.cpp
     MobileGL/MG_Util/Texture/TextureFormatProcessor.cpp
 
@@ -211,6 +230,7 @@ set(SOURCE_FILES
     MobileGL/MG_Impl/GLImpl/Exporting/Definitions.cpp
     MobileGL/MG_Impl/GLImpl/Getter/GL_Getter.cpp
     MobileGL/MG_Impl/GLImpl/Sync/GL_Sync.cpp
+    MobileGL/MG_Impl/GLImpl/Query/GL_Query.cpp
 
     MobileGL/MG_Impl/Init.cpp
     MobileGL/MG_Impl/GetProcAddress.cpp
@@ -238,6 +258,7 @@ set(SOURCE_FILES
     MobileGL/MG_Backend/DirectVulkan/Renderer/VertexInputStateFactory.cpp
     MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferObject.cpp
     MobileGL/MG_Backend/DirectVulkan/Renderer/VkTextureManager.cpp
+    MobileGL/MG_Backend/DirectVulkan/Renderer/VkTimerQueryManager.cpp
     MobileGL/MG_Backend/DirectVulkan/Renderer/VkSamplerManager.cpp
     MobileGL/MG_Backend/DirectVulkan/Renderer/VkClearManager.cpp
     MobileGL/MG_Backend/DirectVulkan/Renderer/VkRenderPassManager.cpp
@@ -270,6 +291,28 @@ set(SOURCE_FILES
     MobileGL/MG_State/GLState/RenderbufferState/RenderbufferState.cpp
 )
 
+if (APPLE AND NOT MOBILEGL_IOS)
+    list(APPEND SOURCE_FILES
+        MobileGL/MG_Impl/CGLImpl/CGLImpl.cpp
+        MobileGL/MG_Impl/CGLImpl/Exporting/Definitions.cpp
+        MobileGL/MG_Impl/DyldInterpose/DyldInterpose.cpp
+        MobileGL/MG_Impl/NSOpenGLImpl/NSOpenGLImpl.cpp
+    )
+endif()
+
+if (ANDROID)
+    list(APPEND SOURCE_FILES
+        MobileGL/MG_Util/SelfTest/DriverPostJni.cpp
+    )
+endif()
+
+if (WIN32)
+    list(APPEND SOURCE_FILES
+        MobileGL/MG_Impl/WGLImpl/WGLImpl.cpp
+        MobileGL/MG_Impl/WGLImpl/Exporting/Definitions.cpp
+    )
+endif()
+
 set(MOBILEGL_LINK_LIBRARIES
         glslang::glslang
         spirv-cross-c
@@ -298,10 +341,18 @@ set(MOBILEGL_INCLUDE_DIR
         ${SPIRV-Headers_SOURCE_DIR}/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
@@ -328,8 +379,34 @@ target_link_libraries(${CMAKE_PROJECT_NAME}
 target_compile_definitions(${CMAKE_PROJECT_NAME}
         PUBLIC
         ${MOBILEGL_COMPILE_DEF}
+        MOBILEGL_LOG_ACTIVE_LEVEL=${MOBILEGL_LOG_ACTIVE_LEVEL}
+        $<$<BOOL:${MOBILEGL_TRACE_ANGLE_VARIANTS}>:MOBILEGL_TRACE_ANGLE_VARIANTS=1>
 )
 
+if(UNIX AND NOT APPLE AND NOT ANDROID)
+    foreach(MOBILEGL_LOADER_ALIAS
+            libEGL.so libEGL.so.1)
+        add_custom_command(TARGET ${CMAKE_PROJECT_NAME} POST_BUILD
+            COMMAND ${CMAKE_COMMAND} -E create_symlink
+                    "$<TARGET_FILE_NAME:${CMAKE_PROJECT_NAME}>"
+                    "$<TARGET_FILE_DIR:${CMAKE_PROJECT_NAME}>/${MOBILEGL_LOADER_ALIAS}"
+            COMMENT "Creating ${MOBILEGL_LOADER_ALIAS} alias for Linux GL/EGL loaders"
+        )
+    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 
         ${SOURCE_FILES}
@@ -361,6 +438,7 @@ if(NOT ANDROID)
     target_compile_definitions(${CMAKE_PROJECT_NAME}_s
             PUBLIC
             ${MOBILEGL_COMPILE_DEF}
+            MOBILEGL_LOG_ACTIVE_LEVEL=${MOBILEGL_LOG_ACTIVE_LEVEL}
     )
 endif()
 
@@ -379,7 +457,48 @@ if (ANDROID)
     )
 endif()
 
-if (NOT ANDROID)
+if (APPLE AND NOT MOBILEGL_IOS)
+    target_link_libraries(${CMAKE_PROJECT_NAME} PUBLIC
+        "-framework Cocoa"
+        "-framework QuartzCore"
+        "-framework Foundation"
+        "-framework OpenGL"
+        objc)
+    if(TARGET ${CMAKE_PROJECT_NAME}_s)
+        target_link_libraries(${CMAKE_PROJECT_NAME}_s PUBLIC
+            "-framework Cocoa"
+            "-framework QuartzCore"
+            "-framework Foundation"
+            "-framework OpenGL"
+            objc)
+    endif()
+endif()
+
+if (APPLE AND MOBILEGL_IOS)
+    target_compile_definitions(${CMAKE_PROJECT_NAME} PUBLIC MOBILEGL_IOS=1 _LIBCPP_DISABLE_AVAILABILITY)
+    target_link_libraries(${CMAKE_PROJECT_NAME} PUBLIC
+        "-framework CoreGraphics"
+        "-framework Foundation"
+        "-framework QuartzCore"
+        objc)
+    if (MOBILEGL_VULKAN_LIBRARY)
+        target_link_libraries(${CMAKE_PROJECT_NAME} PUBLIC "${MOBILEGL_VULKAN_LIBRARY}")
+    endif()
+
+    if(TARGET ${CMAKE_PROJECT_NAME}_s)
+        target_compile_definitions(${CMAKE_PROJECT_NAME}_s PUBLIC MOBILEGL_IOS=1 _LIBCPP_DISABLE_AVAILABILITY)
+        target_link_libraries(${CMAKE_PROJECT_NAME}_s PUBLIC
+            "-framework CoreGraphics"
+            "-framework Foundation"
+            "-framework QuartzCore"
+            objc)
+        if (MOBILEGL_VULKAN_LIBRARY)
+            target_link_libraries(${CMAKE_PROJECT_NAME}_s PUBLIC "${MOBILEGL_VULKAN_LIBRARY}")
+        endif()
+    endif()
+endif()
+
+if (NOT ANDROID AND NOT MOBILEGL_IOS)
     find_package(Vulkan)
     if (Vulkan_FOUND)
         target_link_libraries(${CMAKE_PROJECT_NAME} PUBLIC Vulkan::Vulkan Vulkan::Headers)
@@ -387,6 +506,17 @@ if (NOT ANDROID)
         target_include_directories(${CMAKE_PROJECT_NAME} PUBLIC ${Vulkan_INCLUDE_DIR})
         target_include_directories(${CMAKE_PROJECT_NAME}_s PUBLIC ${Vulkan_INCLUDE_DIR})
     endif ()
+endif ()
+
+if (NOT ANDROID)
+    # Enable testing in the top-level scope so a CTestTestfile.cmake is emitted
+    # at the build-tree root. This lets `ctest` be invoked from the top-level
+    # build directory (IDE "run all tests", CI) and discover every test in the
+    # subdirectories below, instead of having to descend into each
+    # MG_Test/MG_Benchmark subdirectory. Tests are tagged with CTest labels
+    # (unit / benchmark / integration), so e.g. `ctest -L unit` selects just
+    # the unit suite.
+    enable_testing()
 
     if (MOBILEGL_BUILD_TEST)
         add_subdirectory(MobileGL/MG_Test)
@@ -395,4 +525,8 @@ if (NOT ANDROID)
     if (MOBILEGL_BUILD_BENCHMARK)
         add_subdirectory(MobileGL/MG_Benchmark)
     endif()
+
+    if (MOBILEGL_BUILD_TRACE_REPLAY)
+        add_subdirectory(tools/trace_replay)
+    endif()
 endif()

MobileGL/Config.h

@@ -14,9 +14,88 @@ 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, 3, 0, "-dev", VersionType::Development};
+    inline const Version CoreVersion = {26, 7, 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,
+    };
+
+    // 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"
+    // (case-insensitive).
+    //
+    // Env variables intentionally NOT mirrored here (kept as live std::getenv at their
+    // call sites):
+    //   - DISPLAY: X11 session variable, not MobileGL configuration.
+    //   - MOBILEGL_LOG_FILE_PATH: log-file init runs before MG_ConfigLoader::Init
+    //     (see MG_Util/Debug/Log.cpp).
+    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;
+#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_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,
+        // resolves certain rendering bugs on ANGLE + llvmpipe.
+        Bool AvoidSamplerMipmapMinFilter = false;
+        // MOBILEGL_COHERENT_AS_FLUSH: app-compat for engines (e.g. Flywheel) that write
+        // GPU-read data through persistent GL_MAP_FLUSH_EXPLICIT_BIT maps they never
+        // flush. Persistent FLUSH_EXPLICIT map requests are rewritten to coherent
+        // semantics: writes reach the backend without glFlushMappedBufferRange, and
+        // flush calls on rewritten maps become error-free no-ops. Non-persistent maps
+        // keep spec FLUSH_EXPLICIT behavior.
+        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
+        // per-draw glBufferSubData path instead of the persistent-mapped ring allocator
+        // (negative control / driver-bug escape hatch).
+        Bool DisableUboRing = false;
+        // MOBILEGL_RELAXED_SEMANTICS: relax strict core-profile rules (e.g. VAO-0 draws,
+        // texture-name reuse after delete) even on contexts that explicitly requested a core
+        // profile. Without it, relaxed semantics still apply to every context that did not
+        // explicitly request a core profile via EGL_CONTEXT_OPENGL_PROFILE_MASK / a >=3.1
+        // version request.
+        Bool RelaxedSemantics = false;
+        // MOBILEGL_QUIRK_SUBGROUP_PREFIX_SCAN: overrides the shader-source quirk that
+        // rewrites the recognized workgroup prefix-scan template on Qualcomm devices with
+        // subgroups wider than 32 lanes (see ShaderSourceProcessor's quirk registry).
+        QuirkOverride SubgroupPrefixScanQuirk = QuirkOverride::Auto;
+        // MOBILEGL_QUIRK_CLIP_DISTANCE: overrides the DirectGLES quirk that lowers
+        // gl_ClipDistance for Adreno's ESSL compiler (shadow Private arrays with
+        // constant-index builtin flushes, dynamic-index gl_in copy loop, redeclaration
+        // strip, and const struct-array LUT splitting). Auto detects Qualcomm.
+        QuirkOverride ClipDistanceQuirk = QuirkOverride::Auto;
+        // MOBILEGL_MAGMA_DISABLE_BLENDED_DEPTH_WRITE: overrides the DirectVulkan quirk that
+        // strips depth writes from accumulation-blended pipelines (MIN/MAX or additive
+        // ONE+ONE - the multi-pass depth-equality signature) on drivers without
+        // 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_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 DisableRobustBufferAccess = false;
+    };
+    extern FeaturesTable Features;
 } // namespace MobileGL::MG_Config

MobileGL/ConfigLoader.cpp

@@ -8,6 +8,19 @@
 
 #include "Config.h"
 
+#include <cerrno>
+#include <cstdlib>
+
+#ifndef _WIN32
+extern char** environ;
+#endif
+
+namespace MobileGL::MG_Config {
+    // Zero/default-initialized at static-init time (all fields have constexpr-friendly
+    // defaults), so it is safe to read even if MG_ConfigLoader::Init has not run yet.
+    FeaturesTable Features;
+} // namespace MobileGL::MG_Config
+
 namespace MobileGL::MG_ConfigLoader {
     static UniquePtr<UnorderedMap<String, String>> acceptedEnvVariablesMap;
 
@@ -56,6 +69,78 @@ namespace MobileGL::MG_ConfigLoader {
         }
     }
 
+    // Unified truthy rule for boolean feature env variables: set, non-empty, not "0",
+    // and not "false" (case-insensitive).
+    static Bool IsTruthyValue(const String& value) {
+        if (value.empty() || value == "0") {
+            return false;
+        }
+        String lowered = value;
+        std::transform(lowered.begin(), lowered.end(), lowered.begin(),
+                       [](unsigned char c) { return static_cast<char>(std::tolower(c)); });
+        return lowered != "false";
+    }
+
+    inline Bool QueryEnvFlag(const String& key) {
+        auto it = acceptedEnvVariablesMap->find(key);
+        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;
+    }
+
+    inline Uint32 QueryEnvUint32(const String& key, Uint32 defaultValue, Uint32 minValue, Uint32 maxValue) {
+        auto it = acceptedEnvVariablesMap->find(key);
+        if (it == acceptedEnvVariablesMap->end()) {
+            return defaultValue;
+        }
+
+        const String& value = it->second;
+        char* parseEnd = nullptr;
+        errno = 0;
+        const unsigned long parsedValue = std::strtoul(value.c_str(), &parseEnd, 10);
+        if (parseEnd == value.c_str() || *parseEnd != '\0' || errno == ERANGE || parsedValue < minValue ||
+            parsedValue > maxValue) {
+            MGLOG_W("Config: Ignoring invalid env variable %s='%s'; expected an integer in range [%u, %u], "
+                    "using default %u",
+                    key.c_str(), value.c_str(), minValue, maxValue, defaultValue);
+            return defaultValue;
+        }
+
+        return static_cast<Uint32>(parsedValue);
+    }
+
+    inline void InitFeatures() {
+        auto& features = MG_Config::Features;
+        features.DisableTimerQuery = QueryEnvFlag("MOBILEGL_DISABLE_TIMERQUERY");
+        features.UseAngle = QueryEnvFlag("MOBILEGL_USE_ANGLE");
+#if defined(MOBILEGL_TRACE_ANGLE_VARIANTS)
+        QueryEnvVariable("MOBILEGL_TRACE_ANGLE_VARIANT", features.TraceAngleVariant, "");
+#endif
+        features.DisableSubgroup = QueryEnvFlag("MOBILEGL_DISABLE_SUBGROUP");
+        features.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.CoherentAsFlush = QueryEnvFlag("MOBILEGL_COHERENT_AS_FLUSH");
+        features.TraceSkipAutodestroy = QueryEnvFlag("MOBILEGL_TRACE_SKIP_AUTODESTROY");
+        features.DisableUboRing = QueryEnvFlag("MOBILEGL_DISABLE_UBO_RING");
+        features.RelaxedSemantics = QueryEnvFlag("MOBILEGL_RELAXED_SEMANTICS");
+        features.SubgroupPrefixScanQuirk = QueryEnvQuirkOverride("MOBILEGL_QUIRK_SUBGROUP_PREFIX_SCAN");
+        features.ClipDistanceQuirk = QueryEnvQuirkOverride("MOBILEGL_QUIRK_CLIP_DISTANCE");
+        features.MagmaDisableBlendedDepthWriteQuirk =
+            QueryEnvQuirkOverride("MOBILEGL_MAGMA_DISABLE_BLENDED_DEPTH_WRITE");
+        features.DisableRobustBufferAccess = QueryEnvFlag("MOBILEGL_DISABLE_ROBUST_BUFFER_ACCESS");
+    }
+
     inline void InitBackendType() {
         String backendTypeStr;
         QueryEnvVariable("MOBILEGL_BACKEND_TYPE", backendTypeStr, "DirectGLES");
@@ -77,6 +162,7 @@ namespace MobileGL::MG_ConfigLoader {
         InitializeAcceptedEnvVariables();
 
         InitBackendType();
+        InitFeatures();
 
         // Destroy the map since we won't need it anymore
         acceptedEnvVariablesMap.reset();

MobileGL/Defines.h

@@ -32,9 +32,14 @@
 #define MOBILEGL_GLX_API MOBILEGL_API
 #define MOBILEGL_GL_API MOBILEGL_API
 #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 ======================= //
-#define MOBILEGL_LOG_ACTIVE_LEVEL MOBILEGL_LOG_LEVEL_DEBUG
+#ifndef MOBILEGL_LOG_ACTIVE_LEVEL
+#define MOBILEGL_LOG_ACTIVE_LEVEL MOBILEGL_LOG_LEVEL_INFO
+#endif
 
 #define MOBILEGL_LOG_ENABLE_CONSOLE 0
 #define MOBILEGL_LOG_ENABLE_FILE 1
@@ -63,11 +68,15 @@
 #endif
 
 // =============================== Utils ================================ //
-#define MOBILEGL_ASSERT(condition, ...)                                                                                \
-    do {                                                                                                               \
-        if (!(condition)) {                                                                                            \
-            MGLOG_F("Assertion failed" __VA_OPT__(": ") __VA_ARGS__);                                                  \
-            MGLOG_F("  at %s:%d (%s)", __FILE__, __LINE__, __func__);                                                  \
-            TRAP;                                                                                                      \
-        }                                                                                                              \
-    } while (0)
+#if MOBILEGL_LOG_ACTIVE_LEVEL <= MOBILEGL_LOG_LEVEL_DEBUG
+    #define MOBILEGL_ASSERT(condition, ...)                                                                                \
+        do {                                                                                                               \
+            if (!(condition)) {                                                                                            \
+                MGLOG_F("Assertion failed" __VA_OPT__(": ") __VA_ARGS__);                                                  \
+                MGLOG_F("  at %s:%d (%s)", __FILE__, __LINE__, __func__);                                                  \
+                TRAP;                                                                                                      \
+            }                                                                                                              \
+        } while (0)
+#else
+    #define MOBILEGL_ASSERT(condition, ...)
+#endif

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

@@ -32,6 +32,8 @@
 #include <thread>
 #include <vector>
 #include <cassert>
+#include <climits>
+#include <cstdlib>
 #include <cstdarg>
 #include <cstring>
 #include <numeric>
@@ -108,10 +110,32 @@
 #define VK_USE_PLATFORM_WIN32_KHR
 #elif defined(__APPLE__)
 #define VK_USE_PLATFORM_METAL_EXT
+#elif defined(__linux__)
+#define VK_USE_PLATFORM_XLIB_KHR
+typedef struct _XDisplay Display;
+typedef unsigned long XID;
+typedef XID Window;
+typedef unsigned long VisualID;
 #else
 #warning "VK_USE_PLATFORM_*_KHR not defined for this platform!"
 #endif
+#if defined(VK_USE_PLATFORM_XLIB_KHR)
+#pragma push_macro("Bool")
+#pragma push_macro("None")
+#pragma push_macro("Always")
+#pragma push_macro("Status")
+#pragma push_macro("LSBFirst")
+#pragma push_macro("DestroyAll")
+#endif
 #include <vulkan/vulkan.h>
+#if defined(VK_USE_PLATFORM_XLIB_KHR)
+#pragma pop_macro("DestroyAll")
+#pragma pop_macro("LSBFirst")
+#pragma pop_macro("Status")
+#pragma pop_macro("Always")
+#pragma pop_macro("None")
+#pragma pop_macro("Bool")
+#endif
 
 #ifdef TRACY_ENABLE
 #include <tracy/Tracy.hpp>

MobileGL/Init.cpp

@@ -9,13 +9,55 @@
 #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 <atomic>
+#include <mutex>
+
 namespace MobileGL {
+    namespace {
+        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) {
+                return;
+            }
+
+            if (logLifecycle) {
+                MGLOG_I("MobileGL closing...");
+            }
+            glslang::FinalizeProcess();
+            MG_Backend::pActiveBackendObject.reset();
+            MG_State::pGLContext.reset();
+            MG_State::pEGLContext.reset();
+            MG_Impl::GLImpl::TextureImpl::pProxyTextureManager.reset();
+            MG_Impl::GLImpl::FramebufferImpl::pDefaultFramebufferInfo.reset();
+            MG_Backend::gBackendFunctionsTable = {};
+            g_isInitialized = false;
+            if (logLifecycle) {
+                MG_Util::Debug::Close();
+            }
+
+            // TODO: add and use Destroy functions for other subsystems
+        }
+    }
+
     void Initialize() {
+        if (g_isInitialized) {
+            MGLOG_D("MobileGL already initialized; skipping duplicate Initialize()");
+            return;
+        }
+
         MG_Util::Debug::InitFile();
         MGLOG_I("Initializing MobileGL...");
         MG_ConfigLoader::Init();
@@ -28,43 +70,39 @@ namespace MobileGL {
         MGLOG_D("MG_Impl initialized");
         glslang::InitializeProcess();
         MGLOG_D("glslang initialized");
+        g_isInitialized = true;
         MGLOG_I("MobileGL initialized");
     }
 
-    void Destroy() {
-        MGLOG_I("MobileGL closing...");
-        glslang::FinalizeProcess();
-        MG_State::pGLContext.reset();
-        MG_State::pEGLContext.reset();
-        MG_Impl::GLImpl::TextureImpl::pProxyTextureManager.reset();
-        MG_Impl::GLImpl::FramebufferImpl::pDefaultFramebufferInfo.reset();
-        MG_Util::Debug::Close();
-
-        // TODO: add and use Destroy functions for other subsystems
-    }
-
-#if defined(__linux__) || defined(__APPLE__)
-    __attribute__((constructor)) static void AutoInit() {
-        Initialize();
-    }
-
-    __attribute__((destructor)) static void AutoDestroy() {
-        Destroy();
-    }
-#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;
+    void EnsureInitialized() {
+        if (g_isInitialized.load(std::memory_order_acquire)) {
+            return;
         }
-        return TRUE;
+        // 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;
     }
-#endif
+
+    void Destroy() {
+        DestroyImpl(true);
+    }
+
+    // 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 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.
 } // namespace MobileGL

MobileGL/Init.h

@@ -11,6 +11,12 @@
 
 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
+    // MobileGL's lifecycle never depends on ELF/DLL static constructors, and
+    // so a fresh init can follow a full Destroy() (e.g. after the last
+    // eglTerminate).
+    void EnsureInitialized();
     void Destroy();
 
     namespace MG_Util::Debug {

MobileGL/MG_Backend/BackendObject.cpp

@@ -7,18 +7,164 @@
 // End of Source File Header
 
 #include "BackendObject.h"
+#include "MG_Util/Converters/MGToStr/TextureEnumConverter.h"
+#include <algorithm>
+#include <cstring>
+#include <iomanip>
+#include <sstream>
 
 namespace MobileGL::MG_Backend {
     namespace {
         Bool IsReleaseCurrentRequest(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx) {
-            return dpy == EGL_NO_DISPLAY && draw == EGL_NO_SURFACE && read == EGL_NO_SURFACE && ctx == EGL_NO_CONTEXT;
+            (void)dpy;
+            return draw == EGL_NO_SURFACE && read == EGL_NO_SURFACE && ctx == EGL_NO_CONTEXT;
         }
 
         std::thread::id CurrentThreadKey() {
             return std::this_thread::get_id();
         }
+
+        const char* GetFormatCapabilitySupportString(const FormatCapabilityCache& cache,
+                                                     SizeT targetIndex,
+                                                     SizeT formatIndex,
+                                                     FormatCapability capability) {
+            if (HasFormatCapability(cache.FullCaps[targetIndex][formatIndex], capability)) return "Full";
+            if (HasFormatCapability(cache.CaveatCaps[targetIndex][formatIndex], capability)) return "Caveat";
+            return "None";
+        }
+
+        SizeT GetPrintedFormatNameWidth() {
+            SizeT width = 0;
+            for (SizeT formatIndex = 0; formatIndex < kFormatCapabilityFormatCount; ++formatIndex) {
+                const auto format = static_cast<TextureInternalFormat>(formatIndex);
+                width = std::max(width, MG_Util::ConvertTextureInternalFormatToString(format).size());
+            }
+            return width;
+        }
+
+        SizeT GetCapabilityColumnWidth(FormatCapability capability) {
+            SizeT width = std::strlen(GetFormatCapabilityName(capability));
+            width = std::max<SizeT>(width, std::strlen("Caveat"));
+            return width;
+        }
+
+        String BuildFormatCapabilityHeader(SizeT formatNameWidth) {
+            std::ostringstream line;
+            line << std::left << std::setw(static_cast<Int>(formatNameWidth)) << "";
+            for (FormatCapability capability : kReportedFormatCapabilities) {
+                line << " | " << std::left << std::setw(static_cast<Int>(GetCapabilityColumnWidth(capability)))
+                     << GetFormatCapabilityName(capability);
+            }
+            return line.str();
+        }
+
+        String BuildFormatCapabilityRow(const FormatCapabilityCache& cache,
+                                        SizeT targetIndex,
+                                        SizeT formatIndex,
+                                        SizeT formatNameWidth) {
+            const auto format = static_cast<TextureInternalFormat>(formatIndex);
+            std::ostringstream line;
+            line << std::left << std::setw(static_cast<Int>(formatNameWidth))
+                 << MG_Util::ConvertTextureInternalFormatToString(format);
+            for (FormatCapability capability : kReportedFormatCapabilities) {
+                line << " | " << std::left << std::setw(static_cast<Int>(GetCapabilityColumnWidth(capability)))
+                     << GetFormatCapabilitySupportString(cache, targetIndex, formatIndex, capability);
+            }
+            return line.str();
+        }
     } // namespace
 
+    void FormatCapabilityCache::Clear() {
+        for (auto& row : FullCaps) {
+            row.fill(FormatCapabilityFlags{});
+        }
+        for (auto& row : CaveatCaps) {
+            row.fill(FormatCapabilityFlags{});
+        }
+        for (auto& row : SampleCounts) {
+            for (auto& counts : row) {
+                counts.clear();
+            }
+        }
+    }
+
+    Bool HasFormatCapability(FormatCapabilityFlags caps, FormatCapability capability) {
+        return static_cast<Bool>(caps & capability);
+    }
+
+    SizeT GetFormatCapabilityTargetIndex(TextureTarget target) {
+        if (target == TextureTarget::Unknown || static_cast<Int>(target) < 0 ||
+            static_cast<SizeT>(target) >= kFormatCapabilityTextureTargetCount) {
+            return kFormatCapabilityTargetCount;
+        }
+        return static_cast<SizeT>(target);
+    }
+
+    SizeT GetRenderbufferFormatCapabilityTargetIndex() {
+        return kFormatCapabilityRenderbufferTargetIndex;
+    }
+
+    const char* GetFormatCapabilityName(FormatCapability capability) {
+        switch (capability) {
+        case FormatCapability::Creatable:
+            return "Creatable";
+        case FormatCapability::Sampled:
+            return "Sampled";
+        case FormatCapability::LinearFilter:
+            return "LinearFilter";
+        case FormatCapability::GenerateMipmap:
+            return "GenerateMipmap";
+        case FormatCapability::TextureGather:
+            return "TextureGather";
+        case FormatCapability::TextureShadow:
+            return "TextureShadow";
+        case FormatCapability::FramebufferRenderable:
+            return "FramebufferRenderable";
+        case FormatCapability::FramebufferLayered:
+            return "FramebufferLayered";
+        case FormatCapability::MultisampleTexture:
+            return "MultisampleTexture";
+        case FormatCapability::MultisampleRenderbuffer:
+            return "MultisampleRenderbuffer";
+        case FormatCapability::ColorAttachment:
+            return "ColorAttachment";
+        case FormatCapability::DepthAttachment:
+            return "DepthAttachment";
+        case FormatCapability::StencilAttachment:
+            return "StencilAttachment";
+        case FormatCapability::TextureBuffer:
+            return "TextureBuffer";
+        }
+        return "Unknown";
+    }
+
+    String GetFormatCapabilityTargetName(SizeT targetIndex) {
+        if (targetIndex == kFormatCapabilityRenderbufferTargetIndex) {
+            return "Renderbuffer";
+        }
+        if (targetIndex >= kFormatCapabilityTextureTargetCount) {
+            return "Unknown";
+        }
+        return MG_Util::ConvertTextureTargetToString(static_cast<TextureTarget>(targetIndex));
+    }
+
+    void PrintFormatCapabilities(const FormatCapabilityCache& cache) {
+        const SizeT formatNameWidth = GetPrintedFormatNameWidth();
+
+        MGLOG_D("Backend format capabilities:");
+        for (SizeT targetIndex = 0; targetIndex < kFormatCapabilityTargetCount; ++targetIndex) {
+            MGLOG_D("");
+            const String targetName = GetFormatCapabilityTargetName(targetIndex);
+            MGLOG_D("- %s", targetName.c_str());
+            const String header = BuildFormatCapabilityHeader(formatNameWidth);
+            MGLOG_D("%s", header.c_str());
+            for (SizeT formatIndex = 0; formatIndex < kFormatCapabilityFormatCount; ++formatIndex) {
+                const String row = BuildFormatCapabilityRow(cache, targetIndex, formatIndex, formatNameWidth);
+                MGLOG_D("%s", row.c_str());
+            }
+        }
+    }
+
     Bool BackendObject::InitializeEGLDisplay(EGLDisplay dpy, EGLint* major, EGLint* minor) {
         const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
         if (dpy == EGL_NO_DISPLAY) {
@@ -42,28 +188,116 @@ namespace MobileGL::MG_Backend {
         return true;
     }
 
-    Bool BackendObject::CreateEGLWindowSurface(const WindowHandle& handle) {
+    Bool BackendObject::CreateEGLWindowSurface(EGLSurface surface, const WindowHandle& handle) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        return RegisterEGLWindowSurface(surface, handle) && ActivateEGLSurface(surface);
+    }
+
+    Bool BackendObject::ResizeEGLWindowSurface(EGLSurface surface, Uint32 width, Uint32 height) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        auto surfaceIt = m_eglSurfaces.find(surface);
+        if (surfaceIt == m_eglSurfaces.end() || surfaceIt->second.Kind != SurfaceKind::Window) {
+            MGLOG_E("ResizeEGLWindowSurface failed: no window surface is initialized");
+            return false;
+        }
+        surfaceIt->second.Window.Width = width;
+        surfaceIt->second.Window.Height = height;
+        if (m_eglSurface == surface) {
+            m_windowHandle.Width = width;
+            m_windowHandle.Height = height;
+        }
+        return true;
+    }
+
+    Bool BackendObject::CreateEGLPbufferSurface(EGLSurface surface, EGLint width, EGLint height) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        return RegisterEGLPbufferSurface(surface, width, height) && ActivateEGLSurface(surface);
+    }
+
+    Bool BackendObject::RegisterEGLWindowSurface(EGLSurface surface, const WindowHandle& handle) {
         const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
         if (!m_eglDisplayInitialized) {
-            MGLOG_E("CreateEGLWindowSurface failed: EGL display is not initialized");
+            MGLOG_E("RegisterEGLWindowSurface failed: EGL display is not initialized");
+            return false;
+        }
+        if (surface == EGL_NO_SURFACE) {
+            MGLOG_E("RegisterEGLWindowSurface failed: invalid EGLSurface");
             return false;
         }
         if (handle.Backend == WindowBackend::Unknown || !handle.Handle) {
-            MGLOG_E("CreateEGLWindowSurface failed: invalid native window handle");
+            MGLOG_E("RegisterEGLWindowSurface failed: invalid native window handle");
             return false;
         }
 
-        if (m_eglWindowSurfaceInitialized && m_windowHandle.Backend == handle.Backend && m_windowHandle.Handle == handle.Handle) {
+        auto& state = m_eglSurfaces[surface];
+        state = EGLSurfaceState{
+            .Kind = SurfaceKind::Window,
+            .Window = handle,
+            .Width = static_cast<EGLint>(std::max<Uint32>(handle.Width, 1)),
+            .Height = static_cast<EGLint>(std::max<Uint32>(handle.Height, 1)),
+        };
+        return true;
+    }
+
+    Bool BackendObject::RegisterEGLPbufferSurface(EGLSurface surface, EGLint width, EGLint height) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        if (!m_eglDisplayInitialized) {
+            MGLOG_E("RegisterEGLPbufferSurface failed: EGL display is not initialized");
+            return false;
+        }
+        if (surface == EGL_NO_SURFACE) {
+            MGLOG_E("RegisterEGLPbufferSurface failed: invalid EGLSurface");
+            return false;
+        }
+        if (width <= 0 || height <= 0) {
+            MGLOG_E("RegisterEGLPbufferSurface failed: invalid size %dx%d", width, height);
+            return false;
+        }
+
+        m_eglSurfaces[surface] = EGLSurfaceState{
+            .Kind = SurfaceKind::Pbuffer,
+            .Width = width,
+            .Height = height,
+        };
+        return true;
+    }
+
+    const BackendObject::EGLSurfaceState* BackendObject::GetRegisteredEGLSurface(EGLSurface surface) const {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        auto surfaceIt = m_eglSurfaces.find(surface);
+        return surfaceIt == m_eglSurfaces.end() ? nullptr : &surfaceIt->second;
+    }
+
+    Bool BackendObject::ActivateEGLSurface(EGLSurface surface) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        const auto* surfaceState = GetRegisteredEGLSurface(surface);
+        if (!surfaceState) {
+            MGLOG_E("ActivateEGLSurface failed: EGL surface is not registered");
+            return false;
+        }
+        if (m_eglSurfaceInitialized && m_eglSurface == surface) {
             return true;
         }
 
-        SetWindowHandle(handle);
-        if (!InitWindowSurface()) {
-            MGLOG_E("CreateEGLWindowSurface failed: backend InitWindowSurface failed");
+        if (surfaceState->Kind == SurfaceKind::Window) {
+            SetWindowHandle(surfaceState->Window);
+            if (!InitWindowSurface()) {
+                MGLOG_E("ActivateEGLSurface failed: backend InitWindowSurface failed");
+                return false;
+            }
+        } else if (surfaceState->Kind == SurfaceKind::Pbuffer) {
+            if (!InitPbufferSurface(surfaceState->Width, surfaceState->Height)) {
+                MGLOG_E("ActivateEGLSurface failed: backend InitPbufferSurface failed");
+                return false;
+            }
+        } else {
+            MGLOG_E("ActivateEGLSurface failed: unsupported surface kind");
             return false;
         }
 
-        m_eglWindowSurfaceInitialized = true;
+        m_eglSurface = surface;
+        m_eglSurfaceInitialized = true;
+        m_eglSurfaceKind = surfaceState->Kind;
         m_eglCurrentThreads.clear();
         m_backendCapabilitiesInitialized = false;
         return true;
@@ -73,7 +307,7 @@ namespace MobileGL::MG_Backend {
         const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
         const auto threadKey = CurrentThreadKey();
         if (IsReleaseCurrentRequest(dpy, draw, read, ctx)) {
-            m_eglCurrentThreads.erase(threadKey);
+            ReleaseEGLCurrentThread(threadKey);
             return true;
         }
 
@@ -81,8 +315,22 @@ namespace MobileGL::MG_Backend {
             MGLOG_E("MakeEGLCurrent failed: EGL display mismatch or not initialized");
             return false;
         }
-        if (!m_eglWindowSurfaceInitialized) {
-            MGLOG_E("MakeEGLCurrent failed: EGL window surface is not initialized");
+        if (!m_eglSurfaceInitialized) {
+            if (draw != read || !ActivateEGLSurface(draw)) {
+                MGLOG_E("MakeEGLCurrent failed: EGL surface is not initialized");
+                return false;
+            }
+        }
+        if (!GetRegisteredEGLSurface(draw) || !GetRegisteredEGLSurface(read)) {
+            MGLOG_E("MakeEGLCurrent failed: EGL surface is not registered");
+            return false;
+        }
+        if (draw != read) {
+            MGLOG_E("MakeEGLCurrent failed: separate draw/read surfaces are not supported");
+            return false;
+        }
+        if (draw != m_eglSurface && !ActivateEGLSurface(draw)) {
+            MGLOG_E("MakeEGLCurrent failed: EGL surface is not backed by this backend");
             return false;
         }
         if (draw == EGL_NO_SURFACE || read == EGL_NO_SURFACE || ctx == EGL_NO_CONTEXT) {
@@ -98,14 +346,23 @@ namespace MobileGL::MG_Backend {
             m_backendCapabilitiesInitialized = true;
         }
 
-        m_eglCurrentThreads[threadKey] = true;
+        ReleaseEGLCurrentThread(threadKey);
+        m_eglCurrentThreads[threadKey] = EGLCurrentState{
+            .Display = dpy,
+            .DrawSurface = draw,
+            .ReadSurface = read,
+            .Context = ctx,
+        };
         return true;
     }
 
     void BackendObject::ResetEGLRuntimeState() {
         const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
-        m_eglWindowSurfaceInitialized = false;
+        m_eglSurfaceInitialized = false;
         m_backendCapabilitiesInitialized = false;
+        m_eglSurfaceKind = SurfaceKind::None;
+        m_eglSurface = EGL_NO_SURFACE;
+        m_windowHandle = {};
         m_eglCurrentThreads.clear();
     }
 
@@ -115,11 +372,17 @@ namespace MobileGL::MG_Backend {
             MGLOG_E("SwapEGLBuffers failed: EGL display mismatch or not initialized");
             return false;
         }
-        if (m_eglCurrentThreads.find(CurrentThreadKey()) == m_eglCurrentThreads.end()) {
+        const auto currentIt = m_eglCurrentThreads.find(CurrentThreadKey());
+        if (currentIt == m_eglCurrentThreads.end()) {
             MGLOG_E("SwapEGLBuffers failed: no current context attached");
             return false;
         }
-        if (!m_eglWindowSurfaceInitialized || draw == EGL_NO_SURFACE) {
+        if (currentIt->second.Display != dpy || currentIt->second.DrawSurface != draw ||
+            currentIt->second.Context == EGL_NO_CONTEXT) {
+            MGLOG_E("SwapEGLBuffers failed: draw surface is not current on this thread");
+            return false;
+        }
+        if (!m_eglSurfaceInitialized || draw == EGL_NO_SURFACE || draw != m_eglSurface) {
             MGLOG_E("SwapEGLBuffers failed: invalid draw surface");
             return false;
         }
@@ -134,8 +397,99 @@ namespace MobileGL::MG_Backend {
         return true;
     }
 
+    void BackendObject::SetEGLSwapInterval(Int interval) {
+        const auto& backendFunctions = GetBackendFunctions();
+        if (backendFunctions.SetSwapInterval) {
+            backendFunctions.SetSwapInterval(interval);
+        }
+    }
+
+    Bool BackendObject::IsEGLSurfaceCurrent(EGLSurface surface) const {
+        if (surface == EGL_NO_SURFACE) {
+            return false;
+        }
+        for (const auto& current : m_eglCurrentThreads) {
+            if (current.second.DrawSurface == surface || current.second.ReadSurface == surface) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+    void BackendObject::DestroyPendingEGLSurfaceIfUnused(EGLSurface surface) {
+        auto surfaceIt = m_eglSurfaces.find(surface);
+        if (surfaceIt == m_eglSurfaces.end() || !surfaceIt->second.DestroyPending ||
+            IsEGLSurfaceCurrent(surface)) {
+            return;
+        }
+
+        m_eglSurfaces.erase(surfaceIt);
+        if (m_eglSurface == surface) {
+            OnEGLSurfaceReleased(surface);
+            ResetEGLRuntimeState();
+        }
+    }
+
+    void BackendObject::ReleaseEGLCurrentThread(const std::thread::id& threadKey) {
+        auto currentIt = m_eglCurrentThreads.find(threadKey);
+        if (currentIt == m_eglCurrentThreads.end()) {
+            return;
+        }
+
+        const EGLSurface drawSurface = currentIt->second.DrawSurface;
+        const EGLSurface readSurface = currentIt->second.ReadSurface;
+        m_eglCurrentThreads.erase(currentIt);
+        DestroyPendingEGLSurfaceIfUnused(drawSurface);
+        DestroyPendingEGLSurfaceIfUnused(readSurface);
+    }
+
+    void BackendObject::ReleaseEGLSurface(EGLSurface surface) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        auto surfaceIt = m_eglSurfaces.find(surface);
+        if (surfaceIt == m_eglSurfaces.end()) {
+            return;
+        }
+
+        if (IsEGLSurfaceCurrent(surface)) {
+            surfaceIt->second.DestroyPending = true;
+            return;
+        }
+
+        m_eglSurfaces.erase(surfaceIt);
+        if (m_eglSurface == surface) {
+            OnEGLSurfaceReleased(surface);
+            ResetEGLRuntimeState();
+        }
+    }
+
+    void BackendObject::ReleaseEGLResources() {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        ResetEGLRuntimeState();
+        m_eglSurfaces.clear();
+        m_eglDisplay = EGL_NO_DISPLAY;
+        m_eglDisplayInitialized = false;
+    }
+
     void BackendObject::SetWindowHandle(const WindowHandle& handle) {
         const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
         m_windowHandle = handle;
     }
+
+    const FormatCapabilityCache& BackendObject::GetFormatCapabilities() const {
+        return m_formatCapabilities;
+    }
+
+    FormatCapabilityCache& BackendObject::MutableFormatCapabilities() {
+        return m_formatCapabilities;
+    }
+
+    Bool BackendObject::InitPbufferSurface(EGLint width, EGLint height) {
+        (void)width;
+        (void)height;
+        return false;
+    }
+
+    void BackendObject::OnEGLSurfaceReleased(EGLSurface surface) {
+        (void)surface;
+    }
 } // namespace MobileGL::MG_Backend

MobileGL/MG_Backend/BackendObject.h

@@ -8,8 +8,14 @@
 
 #pragma once
 #include <Includes.h>
+#include "MG_State/GLState/TextureState/TextureEnum.h"
 
 namespace MobileGL {
+    namespace MG_State::GLState {
+        class FramebufferObject;
+        class ITextureObject;
+    }
+
     enum class BackendType {
         DirectGLES,
         DirectVulkan,
@@ -18,11 +24,88 @@ namespace MobileGL {
     };
 
     namespace MG_Backend {
+        enum class FormatCapability : Uint64 {
+            Creatable = 1ull << 0,
+
+            Sampled = 1ull << 1,
+            LinearFilter = 1ull << 2,
+            GenerateMipmap = 1ull << 3,
+            TextureGather = 1ull << 4,
+            TextureShadow = 1ull << 5,
+
+            FramebufferRenderable = 1ull << 6,
+            FramebufferLayered = 1ull << 7,
+            MultisampleTexture = 1ull << 8,
+            MultisampleRenderbuffer = 1ull << 9,
+
+            ColorAttachment = 1ull << 10,
+            DepthAttachment = 1ull << 11,
+            StencilAttachment = 1ull << 12,
+
+            TextureBuffer = 1ull << 13
+        };
+
+        using FormatCapabilityFlags = Flags<FormatCapability>;
+
+        inline constexpr Array<FormatCapability, 14> kReportedFormatCapabilities = {
+            FormatCapability::Creatable,
+            FormatCapability::Sampled,
+            FormatCapability::LinearFilter,
+            FormatCapability::GenerateMipmap,
+            FormatCapability::TextureGather,
+            FormatCapability::TextureShadow,
+            FormatCapability::FramebufferRenderable,
+            FormatCapability::FramebufferLayered,
+            FormatCapability::MultisampleTexture,
+            FormatCapability::MultisampleRenderbuffer,
+            FormatCapability::ColorAttachment,
+            FormatCapability::DepthAttachment,
+            FormatCapability::StencilAttachment,
+            FormatCapability::TextureBuffer,
+        };
+
+        inline constexpr SizeT kFormatCapabilityTextureTargetCount =
+            static_cast<SizeT>(TextureTarget::TextureTargetCount);
+        inline constexpr SizeT kFormatCapabilityRenderbufferTargetIndex = kFormatCapabilityTextureTargetCount;
+        inline constexpr SizeT kFormatCapabilityTargetCount = kFormatCapabilityTextureTargetCount + 1;
+        inline constexpr SizeT kFormatCapabilityFormatCount =
+            static_cast<SizeT>(TextureInternalFormat::TextureInternalFormatCount);
+
+        using FormatCapabilityTable =
+            Array<Array<FormatCapabilityFlags, kFormatCapabilityFormatCount>, kFormatCapabilityTargetCount>;
+        using FormatSampleCountTable =
+            Array<Array<Vector<Int>, kFormatCapabilityFormatCount>, kFormatCapabilityTargetCount>;
+
+        struct FormatCapabilityCache {
+            FormatCapabilityTable FullCaps{};
+            FormatCapabilityTable CaveatCaps{};
+            FormatSampleCountTable SampleCounts{};
+
+            void Clear();
+        };
+
+        Bool HasFormatCapability(FormatCapabilityFlags caps, FormatCapability capability);
+        SizeT GetFormatCapabilityTargetIndex(TextureTarget target);
+        SizeT GetRenderbufferFormatCapabilityTargetIndex();
+        const char* GetFormatCapabilityName(FormatCapability capability);
+        String GetFormatCapabilityTargetName(SizeT targetIndex);
+        void PrintFormatCapabilities(const FormatCapabilityCache& cache);
+
+        // Opaque backend fence-sync handle, created by GLFunctionsTable::FenceSync
+        // and released by GLFunctionsTable::DeleteSync.
+        using BackendSyncHandle = void*;
+
+        // Opaque backend timer-query handle, created by
+        // GLFunctionsTable::BeginTimeElapsedQuery / QueryCounterTimestamp and
+        // released by GLFunctionsTable::DeleteBackendQuery.
+        using BackendQueryHandle = void*;
+
         struct GLFunctionsTable {
             void (*DrawArrays)(GLenum mode, GLint first, GLsizei count);
             void (*DrawElements)(GLenum mode, GLsizei count, GLenum type, const void* indices);
             void (*DrawElementsBaseVertex)(GLenum mode, GLsizei count, GLenum type, const void* indices,
                                            GLint basevertex);
+            void (*MultiDrawArrays)(GLenum mode, const GLint* first, const GLsizei* count, GLsizei drawcount);
             void (*MultiDrawElements)(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
                                       GLsizei drawcount);
             void (*MultiDrawElementsBaseVertex)(GLenum mode, const GLsizei* count, GLenum type,
@@ -31,6 +114,10 @@ namespace MobileGL {
             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, GLsizei maxdrawcount, 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,
@@ -54,29 +141,180 @@ namespace MobileGL {
             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 (*ClearNamedFramebufferfi)(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
+                                            GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
             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,
+                                         const SharedPtr<MG_State::GLState::FramebufferObject>& drawFramebuffer,
+                                         GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                                         GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                                         GLbitfield mask, GLenum filter);
             void (*CopyTexImage2D)(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width,
                                    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,
+                                     GLenum srcTarget, GLint srcLevel, GLint srcX, GLint srcY, GLint srcZ,
+                                     const SharedPtr<MG_State::GLState::ITextureObject>& dstTexture,
+                                     GLenum dstTarget, GLint dstLevel, GLint dstX, GLint dstY, GLint dstZ,
+                                     GLsizei srcWidth, GLsizei srcHeight, GLsizei srcDepth);
             void (*GenerateMipmap)(GLenum target);
             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, GLint level, GLenum format, GLenum type,
+                                    GLsizei bufSize, GLvoid* pixels);
+            void (*DispatchCompute)(GLuint numGroupsX, GLuint numGroupsY, GLuint numGroupsZ);
+            void (*DispatchComputeIndirect)(GLintptr indirect);
+            void (*MemoryBarrier)(GLbitfield barriers);
+            void (*MemoryBarrierByRegion)(GLbitfield barriers);
+            void (*BindImageTexture)(GLuint unit, GLuint texture, GLint level, GLboolean layered, GLint layer,
+                                     GLenum access, GLenum format);
+            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);
+            // 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
+            // fence right now (e.g. the calling thread does not own the backend
+            // context); the frontend treats such a sync as always signaled.
+            BackendSyncHandle (*FenceSync)();
+            GLenum (*ClientWaitSync)(BackendSyncHandle sync, GLbitfield flags, GLuint64 timeout);
+            void (*WaitSync)(BackendSyncHandle sync, GLbitfield flags, GLuint64 timeout);
+            void (*DeleteSync)(BackendSyncHandle sync);
+            Bool (*GetSyncStatus)(BackendSyncHandle sync); // true = signaled
+            // GL timer-query objects (GL_ARB_timer_query). All entries are
+            // optional (may be null); the frontend then falls back to zero
+            // results and reports GL_QUERY_COUNTER_BITS == 0.
+            // BeginTimeElapsedQuery / QueryCounterTimestamp may themselves
+            // return null when the backend cannot create a query right now;
+            // the frontend treats such a query as immediately available with
+            // a zero result.
+            // Dynamic support check: true only when the live backend can
+            // actually time at the moment of the call (extension / entry
+            // points / timestamp valid bits are known then, not at table
+            // init). Gates the advertised GL_QUERY_COUNTER_BITS.
+            Bool (*IsTimerQuerySupported)();
+            BackendQueryHandle (*BeginTimeElapsedQuery)();            // starts a TIME_ELAPSED span
+            void (*EndTimeElapsedQuery)(BackendQueryHandle query);    // ends the span
+            BackendQueryHandle (*QueryCounterTimestamp)();            // glQueryCounter(GL_TIMESTAMP) one-shot
+            Bool (*IsQueryResultAvailable)(BackendQueryHandle query); // non-blocking
+            // Returns true when a final value was produced (*outNanoseconds
+            // written; the frontend may cache it and release the handle).
+            // Returns false when the result could not be obtained YET - e.g.
+            // a Vulkan wait that refuses to block on a not-yet-submitted
+            // frame serial - in which case the frontend must keep the handle
+            // and leave the query readable later.
+            Bool (*GetQueryResult64)(BackendQueryHandle query, Bool wait, Uint64* outNanoseconds);
+            void (*DeleteBackendQuery)(BackendQueryHandle query);
+            Int64 (*GetGpuTimestampNs)(); // glGetInteger64v(GL_TIMESTAMP); 0 if unsupported
         };
         struct GlobalBackendFunctionsTable {
             GLFunctionsTable GL;
             void (*Present)();
+            // Optional: applies the app-requested eglSwapInterval to the native
+            // presentation path (null = backend keeps its own pacing policy).
+            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_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;
+            Float AliasedLineWidthRangeMin = 1.0f;
+            Float AliasedLineWidthRangeMax = 1.0f;
+            Float SmoothLineWidthRangeMin = 1.0f;
+            Float SmoothLineWidthRangeMax = 1.0f;
+            Float SmoothLineWidthGranularity = 1.0f;
+            Float PointSizeRangeMin = 1.0f;
+            Float PointSizeRangeMax = 1.0f;
+            Float PointSizeGranularity = 1.0f;
+            Int Max3DTextureSize = 16384;
+            Int MaxArrayTextureLayers = 2048;
+            Int MaxCubeMapTextureSize = 16384;
+            Int MaxFramebufferWidth = 16384;
+            Int MaxFramebufferHeight = 16384;
+            Int MaxFramebufferLayers = 2048;
+            Int MaxRenderbufferSize = 16384;
+            Int MaxTextureSize = 16384;
+            Int MaxColorTextureSamples = 1;
+            Int MaxDepthTextureSamples = 1;
+            Int MaxFramebufferSamples = 1;
+            Int MaxIntegerSamples = 1;
+            Int MaxSamples = 1;
+            Int MaxSampleMaskWords = 1;
+            Int MaxTextureImageUnits = 32;
+            Int MaxVertexTextureImageUnits = 32;
+            Int MaxComputeTextureImageUnits = 32;
+            Int MaxCombinedTextureImageUnits = 192;
+            Int MaxVertexAttribs = 16;
+            Int MaxComputeShaderStorageBlocks = 8;
+            Int MaxCombinedShaderStorageBlocks = 32;
+            Int MaxComputeUniformBlocks = 12;
+            Int MaxComputeWorkGroupInvocations = 128;
+            Int MaxShaderStorageBufferBindings = 8;
+            Int MaxTextureBufferSize = 65536;
+            Int MaxUniformBufferBindings = 24;
+            Int MaxUniformBlockSize = 16384;
+            Int MaxImageUnits = 8;
+            Int MaxCombinedImageUniforms = 8;
+            Int MaxVertexImageUniforms = 0;
+            Int MaxGeometryImageUniforms = 0;
+            Int MaxFragmentImageUniforms = 8;
+            Int MaxComputeImageUniforms = 8;
+            Int MaxDrawBuffers = 8;
+            Int MaxColorAttachments = 8;
+            Int MaxClipDistances = 8;
+            Int MaxViewports = 16;
+            Int MaxViewportWidth = 16384;
+            Int MaxViewportHeight = 16384;
+            Float ViewportBoundsRangeMin = 0.0f;
+            Float ViewportBoundsRangeMax = 0.0f;
+            Int ViewportSubpixelBits = 0;
+            Bool SupportsWideLines = false;
+            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,
-            // TODO: X11, Wayland, Windows, macOS, etc.
+            X11,
+            MetalLayer,
+            Win32, // Handle is an HWND
+            // TODO: Wayland, etc.
             WindowBackendCount,
             Unknown = -1
         };
@@ -84,6 +322,8 @@ namespace MobileGL {
         struct WindowHandle {
             WindowBackend Backend = WindowBackend::Unknown;
             void* Handle = nullptr;
+            Uint32 Width = 0;
+            Uint32 Height = 0;
         };
 
         class BackendObject {
@@ -95,9 +335,16 @@ namespace MobileGL {
             virtual Bool InitWindowSurface() = 0;
 
             virtual Bool InitializeEGLDisplay(EGLDisplay dpy, EGLint* major, EGLint* minor);
-            virtual Bool CreateEGLWindowSurface(const WindowHandle& handle);
+            virtual Bool CreateEGLWindowSurface(EGLSurface surface, const WindowHandle& handle);
+            virtual Bool ResizeEGLWindowSurface(EGLSurface surface, Uint32 width, Uint32 height);
+            virtual Bool CreateEGLPbufferSurface(EGLSurface surface, EGLint width, EGLint height);
             virtual Bool MakeEGLCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx);
             virtual Bool SwapEGLBuffers(EGLDisplay dpy, EGLSurface draw);
+            // Forwards the app-requested eglSwapInterval to the backend's native
+            // presentation path (no-op for backends without a SetSwapInterval hook).
+            virtual void SetEGLSwapInterval(Int interval);
+            virtual void ReleaseEGLSurface(EGLSurface surface);
+            virtual void ReleaseEGLResources();
 
             void SetWindowHandle(const WindowHandle& handle);
 
@@ -105,18 +352,56 @@ namespace MobileGL {
             virtual String GetBackendAPIVersionString() const = 0;
             virtual const GlobalBackendFunctionsTable& GetBackendFunctions() const = 0;
             virtual const DynamicBackendParameters& GetDynamicParameters() const = 0;
+            const FormatCapabilityCache& GetFormatCapabilities() const;
             virtual BackendType GetBackendType() const = 0;
 
         protected:
+            enum class SurfaceKind {
+                None,
+                Window,
+                Pbuffer
+            };
+
+            struct EGLCurrentState {
+                EGLDisplay Display = EGL_NO_DISPLAY;
+                EGLSurface DrawSurface = EGL_NO_SURFACE;
+                EGLSurface ReadSurface = EGL_NO_SURFACE;
+                EGLContext Context = EGL_NO_CONTEXT;
+            };
+
+            struct EGLSurfaceState {
+                SurfaceKind Kind = SurfaceKind::None;
+                Bool DestroyPending = false;
+                WindowHandle Window;
+                EGLint Width = 1;
+                EGLint Height = 1;
+            };
+
             void ResetEGLRuntimeState();
+            Bool RegisterEGLWindowSurface(EGLSurface surface, const WindowHandle& handle);
+            Bool RegisterEGLPbufferSurface(EGLSurface surface, EGLint width, EGLint height);
+            const EGLSurfaceState* GetRegisteredEGLSurface(EGLSurface surface) const;
+            Bool ActivateEGLSurface(EGLSurface surface);
+            virtual Bool InitPbufferSurface(EGLint width, EGLint height);
+            virtual void OnEGLSurfaceReleased(EGLSurface surface);
+            FormatCapabilityCache& MutableFormatCapabilities();
 
             mutable std::recursive_mutex m_eglStateMutex;
+            FormatCapabilityCache m_formatCapabilities;
             WindowHandle m_windowHandle;
             EGLDisplay m_eglDisplay = EGL_NO_DISPLAY;
+            EGLSurface m_eglSurface = EGL_NO_SURFACE;
             Bool m_eglDisplayInitialized = false;
-            Bool m_eglWindowSurfaceInitialized = false;
+            Bool m_eglSurfaceInitialized = false;
             Bool m_backendCapabilitiesInitialized = false;
-            UnorderedMap<std::thread::id, Bool> m_eglCurrentThreads;
+            SurfaceKind m_eglSurfaceKind = SurfaceKind::None;
+            UnorderedMap<std::thread::id, EGLCurrentState> m_eglCurrentThreads;
+            UnorderedMap<EGLSurface, EGLSurfaceState> m_eglSurfaces;
+
+        private:
+            Bool IsEGLSurfaceCurrent(EGLSurface surface) const;
+            void DestroyPendingEGLSurfaceIfUnused(EGLSurface surface);
+            void ReleaseEGLCurrentThread(const std::thread::id& threadKey);
         };
     } // namespace MG_Backend
 } // namespace MobileGL

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

@@ -12,6 +12,12 @@
 #include <MG_Util/BackendLoaders/OpenGL/Loader.h>
 
 namespace MobileGL::MG_Backend::DirectGLES {
+    // Populates the same format-capability cache used by backend startup. The caller
+    // must keep the supplied GLES context current for the duration of this call.
+    void PopulateFormatCapabilities(const MG_External::GLESFunctionsTable& gl,
+                                    const MG_External::GLESCapabilities& capabilities,
+                                    FormatCapabilityCache& cache);
+
     class BackendObject_DirectGLES : public BackendObject {
     public:
         ~BackendObject_DirectGLES() override;
@@ -20,9 +26,12 @@ namespace MobileGL::MG_Backend::DirectGLES {
         Bool InitCapabilities() override;
         Bool InitWindowSurface() override;
         Bool InitializeEGLDisplay(EGLDisplay dpy, EGLint* major, EGLint* minor) override;
-        Bool CreateEGLWindowSurface(const WindowHandle& handle) override;
+        Bool CreateEGLWindowSurface(EGLSurface surface, const WindowHandle& handle) override;
+        Bool CreateEGLPbufferSurface(EGLSurface surface, EGLint width, EGLint height) override;
         Bool MakeEGLCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx) override;
         Bool SwapEGLBuffers(EGLDisplay dpy, EGLSurface draw) override;
+        void ReleaseEGLSurface(EGLSurface surface) override;
+        void ReleaseEGLResources() override;
 
         const RendererInfo& GetRendererInfo() const override;
         String GetBackendAPIVersionString() const override;
@@ -32,9 +41,12 @@ namespace MobileGL::MG_Backend::DirectGLES {
 
         const MG_External::GLESFunctionsTable& GetGLESFunctions() const;
         const MG_External::EGLFunctionsTable& GetEGLFunctions() const;
+        void ApplyGLESCapabilitiesForTesting(const MG_External::GLESCapabilities& capabilities);
 
     private:
         void UpdateDynamicBackendParameters();
+        Bool InitPbufferSurface(EGLint width, EGLint height) override;
+        void OnEGLSurfaceReleased(EGLSurface surface) override;
 
         Bool m_initialized = false;
         MG_External::EGLFunctionsTable m_EGLFunctions;
@@ -42,4 +54,25 @@ namespace MobileGL::MG_Backend::DirectGLES {
         MG_External::GLESCapabilities m_GLESCapabilities;
         DynamicBackendParameters m_dynamicParameters;
     };
+
+    // Single-source-of-truth helpers shared with the driver POST
+    // (MG_Util/SelfTest/DriverPost.cpp), so the identity strings and extension list
+    // MobileGL reports to applications on this backend cannot drift from what the
+    // POST screen shows.
+
+    // Static identity of the Espryt renderer (renderer/backend names, target GL/GLSL
+    // versions, ExtraVendor). The Extensions vector inside is live backend state that
+    // is reconciled after capability init; callers that need the advertised list for
+    // a known capability set must use BuildAdvertisedExtensions instead.
+    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.
+    // The MOBILEGL_DISABLE_TIMERQUERY escape hatch is applied inside.
+    Vector<GLExtension> BuildAdvertisedExtensions(Bool timerQueriesSupported, Bool anisotropicFilteringSupported);
+
+    // Format: <OpenGL ES Renderer>, OpenGL ES <Major>.<Minor> — the exact string an
+    // initialized backend returns from GetBackendAPIVersionString (and that ends up
+    // inside the application-visible GL_RENDERER string).
+    String FormatBackendAPIVersionString(const String& glesRendererString, Int glesMajor, Int glesMinor);
 } // namespace MobileGL::MG_Backend::DirectGLES

MobileGL/MG_Backend/DirectGLES/DirectGLES.h

@@ -8,6 +8,8 @@
 
 #pragma once
 #include <Includes.h>
+#include <MG_Backend/BackendObject.h>
+#include <MG_State/GLState/FramebufferState/FramebufferObject.h>
 #include <MG_State/GLState/TextureState/TextureState.h>
 #include <MG_State/GLState/SamplerState/SamplerObject.h>
 #include <MG_Util/BackendLoaders/OpenGL/Loader.h>
@@ -17,6 +19,10 @@
     operation Utils::CheckGLESError();
 
 namespace MobileGL::MG_Backend::DirectGLES {
+    // Re-establishes the frontend texture-unit bindings on the native ES context.
+    // Content uploads use scratch bindings, so draws and dispatches call this after
+    // texture synchronization.
+    void BindCurrentTextures();
     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);
@@ -25,11 +31,14 @@ namespace MobileGL::MG_Backend::DirectGLES {
     void DrawElements(GLenum mode, GLsizei count, GLenum type, const void* indices);
     void DrawArrays(GLenum mode, GLint first, GLsizei count);
     void DrawElementsBaseVertex(GLenum mode, GLsizei count, GLenum type, const GLvoid* indices, GLint basevertex);
+    void MultiDrawArrays(GLenum mode, const GLint* first, const GLsizei* count, GLsizei drawcount);
     void MultiDrawElements(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
                            GLsizei drawcount);
     void MultiDrawElementsBaseVertex(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
                                      GLsizei drawcount, const GLint* basevertex);
     void MultiDrawElementsIndirect(GLenum mode, GLenum type, const void* indirect, GLsizei drawcount, GLsizei stride);
+    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 DrawRangeElementsBaseVertex(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type,
                                      const void* indices, GLint basevertex);
@@ -46,18 +55,100 @@ namespace MobileGL::MG_Backend::DirectGLES {
                                          GLuint baseinstance);
     void DrawArraysInstanced(GLenum mode, GLint first, GLsizei count, GLsizei instancecount);
     void DrawArraysIndirect(GLenum mode, const void* indirect);
+    void ClearNamedFramebufferfv(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
+                                 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 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,
+                              const SharedPtr<MG_State::GLState::FramebufferObject>& drawFramebuffer,
+                              GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                              GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                              GLbitfield mask, GLenum filter);
     void CopyTexImage2D(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width,
                         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,
+                          GLenum srcTarget, GLint srcLevel, GLint srcX, GLint srcY, GLint srcZ,
+                          const SharedPtr<MG_State::GLState::ITextureObject>& dstTexture,
+                          GLenum dstTarget, GLint dstLevel, GLint dstX, GLint dstY, GLint dstZ,
+                          GLsizei srcWidth, GLsizei srcHeight, GLsizei srcDepth);
     void GenerateMipmap(GLenum target);
     const GLubyte* GetString(GLenum name);
     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 DispatchCompute(GLuint numGroupsX, GLuint numGroupsY, GLuint numGroupsZ);
+    void DispatchComputeIndirect(GLintptr indirect);
+    void MemoryBarrier(GLbitfield barriers);
+    void MemoryBarrierByRegion(GLbitfield barriers);
+    void BindImageTexture(GLuint unit, GLuint texture, GLint level, GLboolean layered, GLint layer, GLenum access,
+                          GLenum format);
+    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);
     Bool InitWindowSurface(NativeWindowType window);
+    Bool InitPbufferSurface(EGLint width, EGLint height);
+    Bool MakeCurrent();
+    Bool ReleaseCurrent();
+    // True when the backend ES context is current on the calling thread, i.e.
+    // immediate buffer ops may issue GL calls right now.
+    Bool IsBackendContextCurrentOnThisThread();
+    // GL fence sync objects, backed by native ES fences. FenceSync returns null
+    // (the frontend then falls back to an always-signaled sync) when the calling
+    // thread does not own the ES context. Waits/queries degrade to "signaled" in
+    // the same situation, and handles created under a since-destroyed ES context
+    // are always treated as signaled.
+    BackendSyncHandle FenceSync();
+    GLenum ClientWaitSync(BackendSyncHandle sync, GLbitfield flags, GLuint64 timeout);
+    void WaitSync(BackendSyncHandle sync, GLbitfield flags, GLuint64 timeout);
+    void DeleteSync(BackendSyncHandle sync);
+    Bool GetSyncStatus(BackendSyncHandle sync);
+    // True when GL_EXT_disjoint_timer_query and every entry point the timer
+    // hooks below need are present. Also gates the E_GL_ARB_timer_query
+    // advertisement in BackendObject_DirectGLES::InitCapabilities, and is
+    // registered as the GLFunctionsTable::IsTimerQuerySupported hook: a pure
+    // capability read needs no current ES context, and it stays false until
+    // the ES capabilities have been filled in.
+    Bool AreTimerQueriesSupported();
+    // 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
+    // context or the extension/entry points are missing, and handles created
+    // under a since-destroyed ES context are always treated as complete with
+    // a zero result (mirrors the fence-sync handles above).
+    BackendQueryHandle BeginTimeElapsedQuery();
+    void EndTimeElapsedQuery(BackendQueryHandle query);
+    BackendQueryHandle QueryCounterTimestamp();
+    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
+    // and release the handle). Returns false only when the calling thread
+    // does not own the ES context, so the value is genuinely unobtainable
+    // right now; the handle stays alive and readable later.
+    Bool GetQueryResult64(BackendQueryHandle query, Bool wait, Uint64* outNanoseconds);
+    void DeleteBackendQuery(BackendQueryHandle query);
+    Int64 GetGpuTimestampNs();
     void Present();
+    // Frame-completion watermarks for the buffer-storage pool: CurrentFrameSerial()
+    // is bumped once per Present(); CompletedFrameSerial() is the newest frame whose
+    // GPU work has provably finished (advanced by polling a one-fence-per-frame ring).
+    // A buffer retired during frame N is safe to recycle once CompletedFrameSerial() >= N.
+    Uint64 CurrentFrameSerial();
+    Uint64 CompletedFrameSerial();
+    // Applies (or defers until the window surface exists) the app-requested
+    // eglSwapInterval on the native EGL surface.
+    void SetSwapInterval(Int interval);
     void SetEGLFuncsTable(const MG_External::EGLFunctionsTable& eglFuncs);
     void SetGLESFuncsTable(const MG_External::GLESFunctionsTable& glesFuncs);
     void SetGLESCapabilities(const MG_External::GLESCapabilities& capabilities);

MobileGL/MG_Backend/DirectGLES/Managers.h

@@ -8,13 +8,19 @@
 
 #pragma once
 #include <Includes.h>
+#include <atomic>
+#include <mutex>
 #include "DirectGLES.h"
 #include "MG_State/GLState/SamplerState/SamplerObject.h"
 #include "MG_State/GLState/TextureState/TextureEnum.h"
 #include <MG_State/GLState/TextureState/TextureObject.h>
 #include <MG_State/GLState/Core.h>
+#include <MG_Util/Converters/MGToGL/TextureEnumConverter.h>
 
 namespace MobileGL::MG_Backend::DirectGLES {
+    String EmulateBaseInstanceInVertexShader(String source, GLenum shaderType);
+    String PromoteDrawParameterGlobalsToUniforms(String source, GLenum shaderType);
+
     template <typename StateObject, typename BackendObject>
     class StateBackendObjectRegistry {
     public:
@@ -50,6 +56,8 @@ namespace MobileGL::MG_Backend::DirectGLES {
             return const_cast<StateBackendObjectRegistry*>(this)->find(stateObj);
         }
 
+        iterator begin() { return m_backendObjects.begin(); }
+        const_iterator begin() const { return m_backendObjects.begin(); }
         iterator end() { return m_backendObjects.end(); }
         const_iterator end() const { return m_backendObjects.end(); }
 
@@ -111,38 +119,150 @@ namespace MobileGL::MG_Backend::DirectGLES {
 
     namespace BufferImpl {
         const GLenum TempBufferTarget = GL_ARRAY_BUFFER;
-        class BackendBufferObject {
+
+        // The DirectGLES storage behind one frontend buffer. Owned (refcounted) by
+        // the frontend BufferObject; immediate BufferBackendOps keep it current, so
+        // draw-time "sync" reduces to ensuring the storage exists.
+        class GLESBufferResource : public MG_State::GLState::BackendBufferResource {
         public:
-            BackendBufferObject();
-            void SyncToBackend(const SharedPtr<MG_State::GLState::BufferObject>& stateBufferObject);
-            Uint GetBackendBufferId() const { return m_backendBufferId; }
-            void Bind(GLenum target = TempBufferTarget);
+            ~GLESBufferResource() override = default;
 
-        private:
-            void SyncToBackend_glBufferData(const SharedPtr<MG_State::GLState::BufferObject>& stateBufferObject);
-            void SyncToBackend_glBufferSubData(const SharedPtr<MG_State::GLState::BufferObject>& stateBufferObject);
-            void SyncToBackend_glMapBufferRange(const SharedPtr<MG_State::GLState::BufferObject>& stateBufferObject,
-                                                Bool invalidate = true, Bool unsynchronized = true);
-
-            Uint m_backendBufferId = 0;
-            SizeT m_prevBufferSize = 0;
-            Bool m_isInitialized = false;
+            Uint id = 0;
+            SizeT storageSize = 0;
+            Bool storageInitialized = false;
+            // ES context generation this resource's id belongs to; ids from a
+            // destroyed context are invalid and must not be deleted or reused.
+            Uint contextGeneration = 0;
+            // Frontend change serial the backend storage reflects. When immediate
+            // ops cannot run (ops unregistered, no current context), this lags and
+            // EnsureBufferResource falls back to a full re-upload. Atomic: read on
+            // the context-owning thread while ops on other threads may update it.
+            std::atomic<Uint64> syncedChangeSerial{0};
+            // Ops that arrived while no ES context was current on the calling thread
+            // (or before storage existed); replayed by EnsureBufferResource. The ES
+            // context migrates between app threads, so deferring ops can race with
+            // the owning thread replaying them: guard both fields with pendingMutex.
+            Bool pendingRespecify = false;
+            VecRange1D pendingRanges;
+            std::mutex pendingMutex;
+            // Zero-copy coherent persistent map (EXT_buffer_storage): the GL store is
+            // immutable, persistently+coherently mapped, and persistentPtr is what the app
+            // (and the frontend PipeResource) write into directly. While set, draw-time
+            // sync is a no-op and no per-draw glBufferSubData is issued. Cleared on ES
+            // context loss.
+            Bool persistentMapped = false;
+            void* persistentPtr = nullptr;
         };
 
-        extern BackendBufferObject* g_boundVertexBufferObject;
-        extern StateBackendObjectRegistry<MG_State::GLState::BufferObject, BackendBufferObject> g_backendBufferObjects;
+        // Registered as the frontend's BufferBackendOps at backend init and on
+        // every MakeCurrent (the ES context can be destroyed and recreated, e.g.
+        // by the trace replayer's probe context).
+        void RegisterBufferBackendOps();
+        void UnregisterBufferBackendOps();
+        // The ES context died: unregister ops, invalidate all outstanding GL ids
+        // (they belonged to the dead context) and drop deferred deletes.
+        void OnBackendContextDestroyed();
+
+        // Get-or-create the backend resource and bring its storage up to date
+        // (creates the GL buffer, replays pending ops, pushes persistent-mapped
+        // ranges). Requires the ES context to be current. Returns nullptr only
+        // for null input.
+        GLESBufferResource* EnsureBufferResource(const SharedPtr<MG_State::GLState::BufferObject>& bufferObject);
+        // Existing resource or nullptr; performs no GL calls.
+        GLESBufferResource* GetBufferResource(MG_State::GLState::BufferObject* bufferObject);
+
+        // Deletes GL buffers whose owning frontend objects died (possibly on a
+        // thread without a current ES context). Called from draw-time sync.
+        void ProcessDeferredBufferReleases();
+
+        // glBindBuffer with a redundant-bind cache for GL_ARRAY_BUFFER.
+        void BindBufferId(GLenum target, Uint id);
+        void InvalidateArrayBufferBindingCache();
+        // Redundant-bind caches for the driver-level GL_PIXEL_PACK/UNPACK_BUFFER
+        // bindings. Every backend readback (glReadPixels / pack-PBO map) and pixel
+        // upload site routes its binding through these so the shadow always matches
+        // the driver; the resting state between operations is 0, which keeps any
+        // path that implicitly assumes "no PBO bound" correct. Scrubbed when a
+        // buffer id is deleted/pooled (GL resets a deleted buffer's bindings to 0,
+        // and a recycled name matching the shadow would false-skip the rebind) and
+        // invalidated on MakeCurrent (context may reset).
+        void BindPixelPackBufferId(Uint id);
+        void BindPixelUnpackBufferId(Uint id);
+        void InvalidatePixelBufferBindingCaches();
+        // A GL buffer id is being deleted by code outside BufferImpl (e.g. the VAO
+        // client-attribute staging buffers): scrub every buffer-binding shadow that
+        // could false-skip when the name is recycled.
+        void NoteBufferIdDeleted(Uint id);
+        // Redundant-bind cache for INDEXED buffer bindings (glBindBufferBase/Range on
+        // GL_UNIFORM_BUFFER / GL_SHADER_STORAGE_BUFFER): skips the GL call when the
+        // (id, range) already at that index matches, like the array-buffer/texture/
+        // sampler caches already do. Invalidated on MakeCurrent (context may reset).
+        void BindBufferBaseCached(GLenum glTarget, Uint index, Uint id);
+        void BindBufferRangeCached(GLenum glTarget, Uint index, Uint id, GLintptr offset, GLsizeiptr size);
+        void InvalidateIndexedBufferBindingCache();
+        // Buffer-storage pool maintenance. TrimBufferPool evicts over-budget entries
+        // (called once per frame from Present); ClearBufferPool drops all pooled ids
+        // without glDeleteBuffers (called when the ES context is going away).
+        void TrimBufferPool();
+        void ClearBufferPool();
+
+        // --- Global-UBO ring ------------------------------------------------------
+        // One persistently+coherently mapped buffer (EXT_buffer_storage) shared by
+        // every program's lowered default-uniform block. Each content change is
+        // bump-allocated into a fresh slot and bound with glBindBufferRange, so the
+        // CPU never rewrites bytes the GPU may still be reading — the per-draw
+        // glBufferSubData into one static UBO forced Adreno to resolve that
+        // write-after-read hazard on every uniform-dirtying draw (MC dirties
+        // uniforms every draw). Reclamation rides the Present() frame-fence
+        // watermark; no ring bytes are recycled before their frame's GPU work
+        // completed.
+        //
+        // A program's cached slot, reusable within one frame while the frontend UBO
+        // content version is unchanged. Cross-frame reuse is intentionally not
+        // attempted: later same-frame allocations may recycle bytes of completed
+        // frames, so re-referencing them would need per-bind pinning — rewriting
+        // GetUBOSize() bytes once per program per frame is far cheaper.
+        struct UboRingAllocation {
+            Uint32 contentVersion = ~0u; // frontend UBO content version held at `offset`
+            Uint32 ringGeneration = 0;   // ring identity the slot lives in (0 = never valid)
+            Uint64 frameSerial = ~Uint64{0}; // frame the slot was written in
+            SizeT offset = 0;
+        };
+        // False when the feature is disabled, EXT_buffer_storage / fences are
+        // missing, the ES context is not current, or ring creation already failed
+        // under this context (callers then take the legacy glBufferSubData path).
+        Bool UboRingAvailable();
+        // Bump-allocate `size` bytes aligned to GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT.
+        // Grows the ring (new GL store, generation bump) when the in-flight span
+        // would be overrun. Returns false when storage (re)creation fails.
+        Bool UboRingAllocate(SizeT size, SizeT& outOffset);
+        void* UboRingMappedPtr();
+        Uint UboRingBufferId();
+        Uint32 UboRingGeneration();
+        // Present()-time upkeep: records the frame's high-water mark for reclamation
+        // and deletes grown-away ring stores once the GPU is done with them.
+        void UboRingOnPresent();
     } // namespace BufferImpl
 
     namespace VertexArrayImpl {
         class BackendVertexArrayObject {
         public:
             BackendVertexArrayObject();
+            ~BackendVertexArrayObject();
             void SyncToBackend(const SharedPtr<MG_State::GLState::VertexArrayObject>& stateVAOObject);
+            void SyncClientSideAttributesForDrawArrays(
+                const SharedPtr<MG_State::GLState::VertexArrayObject>& stateVAOObject, GLint first, GLsizei count);
             Uint GetBackendVertexArrayId() const { return m_backendVAOId; }
             void Bind() const;
 
         private:
             Uint m_backendVAOId = 0;
+            Array<Uint, MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS> m_clientAttributeBufferIds;
+            // Attribs the frontend has Enabled but that have no source at all (no buffer object
+            // and NULL client pointer). GL keeps such attribs latently enabled, but Adreno's ES
+            // driver treats them as client arrays and memcpys from address 0 at draw time
+            // (SIGSEGV), so they are kept disabled on the backend VAO until they gain a source.
+            Uint32 m_forceDisabledAttribsMask = 0;
             Bool m_isInitialized = false;
             Uint16 m_syncedIndexBufferVersion = 0;
             Array<MG_State::GLState::VertexAttributeVersion, MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS>
@@ -155,11 +275,57 @@ namespace MobileGL::MG_Backend::DirectGLES {
 
     namespace TextureImpl {
         inline Bool IsSupportedTextureTarget(TextureTarget target) {
-            if (target == TextureTarget::Texture1D || target == TextureTarget::TextureRectangle ||
-                target == TextureTarget::Texture2DMultisampleArray || target == TextureTarget::Texture1DArray ||
-                target == TextureTarget::Texture2DMultisample || target == TextureTarget::Texture2DArray)
-                return false;
-            return true;
+            // Rectangle textures need non-normalized sampling ES cannot express; everything else is
+            // either native or emulated (1D -> 2D with height 1, 1D array -> 2D array, see
+            // MapToBackendTextureTarget). SPIRV-Cross already emits the matching ESSL samplers and
+            // coordinate padding for 1D/1D-array shaders.
+            return target != TextureTarget::TextureRectangle;
+        }
+
+        // ES has no 1D targets: 1D textures are stored as 2D (height 1) and 1D arrays as 2D arrays
+        // (height 1, layers in depth). Must match SPIRV-Cross's ES 1D-as-2D shader emulation.
+        inline TextureTarget MapToBackendTextureTarget(TextureTarget target) {
+            switch (target) {
+            case TextureTarget::Texture1D:
+                return TextureTarget::Texture2D;
+            case TextureTarget::Texture1DArray:
+                return TextureTarget::Texture2DArray;
+            default:
+                return target;
+            }
+        }
+
+        inline GLenum ConvertTextureTargetToBackendGLEnum(TextureTarget target) {
+            return MG_Util::ConvertTextureTargetToGLEnum(MapToBackendTextureTarget(target));
+        }
+
+        inline GLenum ConvertTextureUploadTargetToBackendGLEnum(TextureUploadTarget uploadTarget) {
+            switch (uploadTarget) {
+            case TextureUploadTarget::Texture1D:
+                return GL_TEXTURE_2D;
+            case TextureUploadTarget::Texture1DArray:
+                return GL_TEXTURE_2D_ARRAY;
+            default:
+                return MG_Util::ConvertTextureUploadTargetToGLEnum(uploadTarget);
+            }
+        }
+
+        // 1D arrays store layers in the state-side height; the ES 2D-array image keeps height 1 and
+        // moves the layer count into depth.
+        inline IntVec3 GetBackendUploadSize(TextureTarget stateTarget, const IntVec3& texelSize) {
+            if (stateTarget == TextureTarget::Texture1DArray) {
+                return {texelSize.x(), 1, texelSize.y()};
+            }
+            return texelSize;
+        }
+
+        inline Bool IsMultisampleTextureTarget(TextureTarget target) {
+            return target == TextureTarget::Texture2DMultisample ||
+                   target == TextureTarget::Texture2DMultisampleArray;
+        }
+
+        inline Bool SupportsWrapR(TextureTarget target) {
+            return target == TextureTarget::Texture3D || target == TextureTarget::TextureCubeMap;
         }
 
         struct StateTextureBasicInfo { // Used for tracking texture state changes
@@ -169,11 +335,14 @@ namespace MobileGL::MG_Backend::DirectGLES {
             SizeT depth = 0;
             SizeT mipmapLevels = 0;
             Uint bufferExternalIndex = 0;
+            Int samples = 0;
+            Bool fixedSampleLocations = true;
 
             bool operator==(const StateTextureBasicInfo& other) const {
                 return internalFormat == other.internalFormat && width == other.width && height == other.height &&
                        depth == other.depth && mipmapLevels == other.mipmapLevels &&
-                       bufferExternalIndex == other.bufferExternalIndex;
+                       bufferExternalIndex == other.bufferExternalIndex && samples == other.samples &&
+                       fixedSampleLocations == other.fixedSampleLocations;
             }
 
             bool operator!=(const StateTextureBasicInfo& other) const { return !(*this == other); }
@@ -183,15 +352,29 @@ namespace MobileGL::MG_Backend::DirectGLES {
         class BackendTextureObject {
         public:
             BackendTextureObject();
+            // Deletes the GL texture (frontend glDeleteTextures used to leak every
+            // backend id for the context lifetime) and scrubs the binding/scratch-FBO
+            // shadows so a recycled name or heap address cannot false-skip a rebind.
+            ~BackendTextureObject();
+            BackendTextureObject(const BackendTextureObject&) = delete;
+            BackendTextureObject& operator=(const BackendTextureObject&) = delete;
             void SyncMipmapsToBackend(const SharedPtr<MG_State::GLState::ITextureObject>& stateTextureObject);
             void SyncBuiltinSamplerToBackend(const SharedPtr<MG_State::GLState::ITextureObject>& stateTextureObject);
             void SyncTextureParamsToBackend(const SharedPtr<MG_State::GLState::ITextureObject>& stateTextureObject);
+            void RequireImageBindableStorage();
             void Bind(GLenum target, Uint unit = TempTextureUnit);
             Uint GetBackendTextureId() const;
 
         private:
+            void RecreateBackendTexture();
+
             Uint m_backendTextureId = 0;
+            // ES context generation the id was created under; a dtor running after
+            // that context died must not delete a foreign (recycled) name.
+            Uint m_contextGeneration = 0;
             Bool m_isInitialized = false;
+            Bool m_imageBindableStorageRequired = false;
+            Bool m_backendStorageImmutable = false;
             StateTextureBasicInfo m_prevTextureInfo;
             SamplerParameters m_cacheSamplerParameters;
             UintVec2 m_cacheLodRange = {0, 1000};
@@ -206,10 +389,16 @@ namespace MobileGL::MG_Backend::DirectGLES {
         void UnbindTexture(Uint unit, GLenum target);
         extern StateBackendObjectRegistry<MG_State::GLState::ITextureObject, BackendTextureObject>
             g_backendTextureObjects;
+        SharedPtr<BackendTextureObject>& SyncTextureObjectToBackend(
+            const SharedPtr<MG_State::GLState::ITextureObject>& textureObject,
+            Bool imageBindableStorageRequired = false);
         extern Array<Array<BackendTextureObject*, (SizeT)TextureTarget::TextureTargetCount>,
                      MG_State::GLState::TextureState::MAX_TEXTURE_IMAGE_UNITS>
             g_boundTexturesCache;
         extern Uint g_activeTextureUnit;
+        // Bumped when the backend ES context is destroyed; texture ids stamped with
+        // an older generation belong to a dead context and must not be deleted.
+        extern Uint g_textureContextGeneration;
     } // namespace TextureImpl
 
     namespace FramebufferImpl {
@@ -218,6 +407,11 @@ namespace MobileGL::MG_Backend::DirectGLES {
             BackendFramebufferObject();
             void SyncToBackend(const SharedPtr<MG_State::GLState::FramebufferObject>& stateFBOObject,
                                FramebufferTarget asTarget);
+            // Apply only this FBO's read buffer (glReadBuffer) to the backend. Split out so it can
+            // still run when SyncCurrentFBO skips the READ-target sync because the same GL FBO is
+            // bound as both draw and read (otherwise glReadBuffer changes would be silently dropped).
+            void SyncReadBufferToBackend(const SharedPtr<MG_State::GLState::FramebufferObject>& stateFBOObject);
+            void InvalidateSyncedState();
             Uint GetBackendFramebufferId() const { return m_backendFBOId; }
             void Bind(FramebufferTarget target) const;
             //            FramebufferAttachmentType GetCompactedAttachmentTypeAtDrawBufferIndex(Int index);
@@ -250,24 +444,200 @@ namespace MobileGL::MG_Backend::DirectGLES {
         extern StateBackendObjectRegistry<MG_State::GLState::FramebufferObject, BackendFramebufferObject>
             g_backendFramebufferObjects;
         extern Array<Uint16, SizeT(FramebufferTarget::FramebufferTargetCount)> g_fboBindVersions;
+        // Tracks the bound FBO's object version (bumped on any attachment/drawbuffer change)
+        // per target: re-attaching textures or changing draw buffers on an already-bound FBO
+        // must re-sync it even when the binding-slot version has not moved.
+        extern Array<Uint16, SizeT(FramebufferTarget::FramebufferTargetCount)> g_fboSyncedObjectVersions;
+        extern Array<MG_State::GLState::FramebufferObject*, SizeT(FramebufferTarget::FramebufferTargetCount)>
+            g_fboSyncedObjects;
+
+        // Driver-level READ/DRAW framebuffer-binding shadow. Every backend
+        // glBindFramebuffer routes through BindFramebufferId so scoped helpers can
+        // save/restore the current binding without a glGetIntegerv round-trip (that
+        // query forces a driver pipeline sync) and so redundant rebinds no-op.
+        // Starts unknown; the first CurrentFramebufferBinding() query pins it from
+        // the driver once. Invalidated on MakeCurrent (context may reset).
+        // GL_FRAMEBUFFER binds both targets.
+        void BindFramebufferId(GLenum fbTarget, Uint id);
+        Uint CurrentFramebufferBinding(FramebufferTarget target);
+        void InvalidateFramebufferBindingCache();
     } // namespace FramebufferImpl
 
+    // Shared scratch framebuffers for the readback/copy/blit emulation paths, with a
+    // driver-side attachment shadow: repeated uses skip redundant detach/attach GL
+    // calls, and an attachment left by one use (e.g. a depth copy's DEPTH_STENCIL
+    // texture) is detached exactly when a later use of another aspect would
+    // otherwise inherit it (stale cross-aspect attachments made the shared temp FBO
+    // incomplete and silently degraded later readbacks).
+    namespace ScratchFBOImpl {
+        struct ScratchFramebuffer {
+            Uint id = 0;
+            // false => attachment state unknown; scrub every point on next use.
+            // A fresh FBO starts with nothing attached, so creation sets it true.
+            Bool attachmentsKnown = false;
+            Uint colorTex = 0;
+            GLenum colorTarget = 0;
+            GLint colorLevel = 0;
+            GLint colorLayer = -1; // >= 0 => attached via glFramebufferTextureLayer
+            Uint depthTex = 0;
+            GLenum depthTarget = 0;
+            GLint depthLevel = 0;
+            Bool depthHasStencil = false;
+            // Per-FBO read/draw buffer state (0 = unknown, set on first use).
+            GLenum readBuffer = 0;
+            GLenum drawBuffer = 0;
+        };
+        ScratchFramebuffer& TempFramebuffer();     // GetTexImage READ / CopyTex*Image2D depth DRAW
+        ScratchFramebuffer& BlitReadFramebuffer(); // texture-to-texture blit source
+        ScratchFramebuffer& BlitDrawFramebuffer(); // texture-to-texture blit destination
+        // Returns the GL id, generating it if needed (requires a current ES context).
+        Uint EnsureId(ScratchFramebuffer& fb);
+        // The fb must currently be bound at fbTarget (glReadBuffer/glDrawBuffers
+        // target the READ/DRAW binding respectively). Each Ensure* performs the
+        // minimal detach/attach set and keeps the shadow in sync; a failed attach
+        // records the point as detached so the completeness check fails instead of
+        // silently reading a stale attachment.
+        void EnsureColorAttachment2D(ScratchFramebuffer& fb, GLenum fbTarget, Uint tex, GLenum texTarget, GLint level);
+        void EnsureColorAttachmentLayer(ScratchFramebuffer& fb, GLenum fbTarget, Uint tex, GLint level, GLint layer);
+        void EnsureDepthAttachment2D(ScratchFramebuffer& fb, GLenum fbTarget, Uint tex, GLenum texTarget, GLint level,
+                                     Bool withStencil);
+        void EnsureNoColorAttachment(ScratchFramebuffer& fb, GLenum fbTarget);
+        void EnsureNoDepthAttachment(ScratchFramebuffer& fb, GLenum fbTarget);
+        void EnsureReadBuffer(ScratchFramebuffer& fb, GLenum readBuffer);
+        void EnsureDrawBuffer(ScratchFramebuffer& fb, GLenum drawBuffer);
+        // A 1x1 RGBA8-renderbuffer-complete FBO (GenerateMipmap needs a complete
+        // binding while respecifying texture storage). Attachment is set once at
+        // creation and never changes.
+        Uint EnsureCompleteTinyFramebufferId();
+        // A backend texture id is being deleted or respecified: a scratch FBO still
+        // referencing it would hold a dangling attachment (ES only auto-detaches
+        // from the *bound* framebuffer), and a recycled name could false-skip a
+        // re-attach; force a full scrub on next use.
+        void NoteTextureIdDeleted(Uint textureId);
+        // The ES context (and the scratch FBO ids with it) is going away.
+        void OnBackendContextDestroyed();
+    } // namespace ScratchFBOImpl
+
+    // Driver-level GL_PACK_* pixel-store shadow, the readback-side sibling of the
+    // upload path's ScopedDefaultUnpackState (Managers.cpp): the backend PACK state
+    // is written ONLY through ApplyPackState, so scoped helpers can save/restore it
+    // from the shadow instead of glGetIntegerv (which forces a driver pipeline
+    // sync), and redundant glPixelStorei calls no-op. The first Apply/Current call
+    // pins the driver to the shadow by writing all fields once. Invalidated on
+    // MakeCurrent (context may reset). PACK_IMAGE_HEIGHT/SKIP_IMAGES/SWAP_BYTES/
+    // LSB_FIRST have no ES equivalents; readbacks honor them on the CPU from the
+    // frontend context state instead.
+    namespace PixelStoreImpl {
+        struct PackState {
+            GLint Alignment = 4;
+            GLint RowLength = 0;
+            GLint SkipRows = 0;
+            GLint SkipPixels = 0;
+            Bool operator==(const PackState& o) const {
+                return Alignment == o.Alignment && RowLength == o.RowLength && SkipRows == o.SkipRows &&
+                       SkipPixels == o.SkipPixels;
+            }
+        };
+        void ApplyPackState(const PackState& desired);
+        PackState CurrentPackState();
+        void InvalidatePackStateCache();
+    } // namespace PixelStoreImpl
+
+    // Image uniforms take their unit from the layout(binding=N) qualifier baked into
+    // the transpiled ESSL; unlike samplers they must not (and in ES cannot) be
+    // assigned through glUniform1i.
+    inline Bool IsImageUniformType(GLenum type) {
+        switch (type) {
+        case 0x904D: /*GL_IMAGE_2D*/
+        case 0x904E: /*GL_IMAGE_3D*/
+        case 0x9050: /*GL_IMAGE_CUBE*/
+        case 0x9051: /*GL_IMAGE_BUFFER*/
+        case 0x9053: /*GL_IMAGE_2D_ARRAY*/
+        case 0x9058: /*GL_INT_IMAGE_2D*/
+        case 0x9059: /*GL_INT_IMAGE_3D*/
+        case 0x905B: /*GL_INT_IMAGE_CUBE*/
+        case 0x905C: /*GL_INT_IMAGE_BUFFER*/
+        case 0x905E: /*GL_INT_IMAGE_2D_ARRAY*/
+        case 0x9063: /*GL_UNSIGNED_INT_IMAGE_2D*/
+        case 0x9064: /*GL_UNSIGNED_INT_IMAGE_3D*/
+        case 0x9066: /*GL_UNSIGNED_INT_IMAGE_CUBE*/
+        case 0x9067: /*GL_UNSIGNED_INT_IMAGE_BUFFER*/
+        case 0x9069: /*GL_UNSIGNED_INT_IMAGE_2D_ARRAY*/
+            return true;
+        default:
+            return false;
+        }
+    }
+
     namespace PrgramImpl {
         class BackendProgramObjectImpl {
         public:
+            // Per-link cache of a sampler-style uniform's backend location: built once in
+            // SyncToBackend so draws stop issuing glGetUniformLocation string queries.
+            // lastAssignedUnit mirrors the program-state value set through glUniform1i
+            // (program state persists across binds, so caching per program is exact).
+            struct SamplerUniformBinding {
+                Uint frontendLocation = 0;
+                Int backendLocation = -1;
+                GLenum uniformType = 0;
+                Int lastAssignedUnit = -1;
+            };
+
             BackendProgramObjectImpl();
             ~BackendProgramObjectImpl();
             void SyncToBackend(const SharedPtr<MG_State::GLState::ProgramObject>& stateProgramObject);
             void Use() const;
+            void SetBaseInstance(Uint32 baseInstance) const;
+            void SetBaseInstanceWordIndex(Int32 wordIndex) const;
+            void SetDrawID(Uint32 drawId) const;
+            Int GetIndirectParamsBinding() const { return m_indirectParamsBinding; }
             Uint GetBackendProgramId() const { return m_backendProgramId; }
             Uint GetBackendGlobalUBOId() const { return m_backendGlobalUBOId; }
+            Uint32 GetSnormFallbackClampOutputMask() const { return m_snormFallbackClampOutputMask; }
+            Uint32 GetUnormFallbackClampOutputMask() const { return m_unormFallbackClampOutputMask; }
+
+            Bool HasGlobalUboBlock() const { return m_globalUboBackendBlockIndex >= 0; }
+            const Vector<Int>& GetUniformBlockBackendIndices() const { return m_uniformBlockBackendIndices; }
+            Vector<SamplerUniformBinding>& GetSamplerUniformBindings() { return m_samplerUniformBindings; }
+            Uint32 GetLastUploadedGlobalUboVersion() const { return m_lastUploadedGlobalUboVersion; }
+            void SetLastUploadedGlobalUboVersion(Uint32 version) { m_lastUploadedGlobalUboVersion = version; }
+            // Backend-reported GL_UNIFORM_BLOCK_DATA_SIZE of the global block; ring
+            // bindings must span at least this much (may exceed the frontend's
+            // reflected size when the transpiled block pads differently).
+            Int GetGlobalUboBackendBlockSize() const { return m_globalUboBackendBlockSize; }
+            BufferImpl::UboRingAllocation& GetGlobalUboRingAllocation() { return m_globalUboRingAllocation; }
+            // Frontend link version this backend program (and its resource caches) was
+            // built from; a mismatch means every link-derived cache here is stale.
+            Uint32 GetSyncedLinkVersion() const { return m_syncedLinkVersion; }
 
         private:
+            void CacheResourceLocations(const SharedPtr<MG_State::GLState::ProgramObject>& stateProgramObject);
+
             Uint m_backendProgramId = 0;
             Uint m_backendGlobalUBOId = 0;
+            Int m_baseInstanceUniformLocation = -1;
+            Int m_drawIdUniformLocation = -1;
+            Int m_baseInstanceWordIndexUniformLocation = -1;
+            Int m_indirectParamsBinding = -1;
+            Uint32 m_snormFallbackClampOutputMask = 0;
+            Uint32 m_unormFallbackClampOutputMask = 0;
             Bool m_isInitialized = false;
+
+            Int m_globalUboBackendBlockIndex = -1;
+            Int m_globalUboBackendBlockSize = 0;
+            Vector<Int> m_uniformBlockBackendIndices; // frontend block index -> backend index (-1 = absent)
+            Vector<SamplerUniformBinding> m_samplerUniformBindings;
+            Uint32 m_lastUploadedGlobalUboVersion = ~0u;
+            BufferImpl::UboRingAllocation m_globalUboRingAllocation;
+            Uint32 m_syncedLinkVersion = ~0u;
         };
 
+        extern Uint32 g_snormFallbackClampOutputMask;
+        extern Uint32 g_unormFallbackClampOutputMask;
+        // Backend id of the last glUseProgram issued through this backend; lets Use()
+        // skip redundant rebinds. Reset to 0 wherever glUseProgram(0) is issued or the
+        // ES context is recreated.
+        extern Uint g_lastUsedBackendProgramId;
         extern StateBackendObjectRegistry<MG_State::GLState::ProgramObject, BackendProgramObjectImpl>
             g_backendProgramObjects;
     } // namespace PrgramImpl
@@ -309,6 +679,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
             TextureInternalFormat m_cacheInternalFormat = TextureInternalFormat::Unknown;
             Int m_cacheWidth = 0;
             Int m_cacheHeight = 0;
+            Int m_cacheSamples = 0;
         };
 
         extern StateBackendObjectRegistry<MG_State::GLState::RenderbufferObject, BackendRenderbufferObject>

MobileGL/MG_Backend/DirectGLES/Utils.cpp

@@ -9,6 +9,7 @@
 #include "DirectGLES.h"
 #include "Utils.h"
 #include "Managers.h"
+#include "MG_Backend/BackendObjects.h"
 #include "MG_Util/Converters/GLToMG/FramebufferEnumConverter.h"
 #include "MG_Util/Texture/TextureFormatProcessor.h"
 
@@ -17,19 +18,135 @@
 #include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
 #include <MG_Util/Converters/MGToGL/TextureEnumConverter.h>
 #include <MG_Util/Converters/MGToGL/FramebufferEnumConverter.h>
+#include <MG_Util/Math/HalfFloat.h>
+#include <MG_Util/Math/SmallFloat.h>
+
+#include <cmath>
 
 namespace MobileGL::MG_Backend::DirectGLES {
+    namespace {
+        Flags<PixelFormatNormalizeOptionBit> GetForcedPixelFormatNormalizeOptions() {
+            Flags<PixelFormatNormalizeOptionBit> options;
+            if (g_GLESCapabilities.IsAngleRenderer) {
+                options |= PixelFormatNormalizeOptionBit::NoRgb16;
+                options |= PixelFormatNormalizeOptionBit::NoSnorm16;
+                options |= PixelFormatNormalizeOptionBit::NoSnorm8;
+            }
+            return options;
+        }
+
+        Flags<PixelFormatNormalizeOptionBit> GetDriverPixelFormatNormalizeOptions() {
+            Flags<PixelFormatNormalizeOptionBit> options = PixelFormatNormalizeOptionBit::NoDepthComponent32;
+            options |= PixelFormatNormalizeOptionBit::NoRGBA8Snorm;
+            options |= PixelFormatNormalizeOptionBit::NoRGB16Snorm;
+            if (!g_GLESCapabilities.SupportsNorm16Texture) {
+                options |= PixelFormatNormalizeOptionBit::NoNorm16;
+            }
+            return options;
+        }
+
+        Flags<PixelFormatNormalizeOptionBit> GetRuntimeFallbackNormalizeOptions(GLenum requestedInternalFormat) {
+            using namespace MG_Util::TextureFormatProcessor;
+            const Flags<PixelFormatNormalizeOptionBit> forcedOptions =
+                GetApplicablePixelFormatNormalizeOptions(requestedInternalFormat, GetForcedPixelFormatNormalizeOptions());
+            if (forcedOptions) {
+                return forcedOptions;
+            }
+            return GetApplicablePixelFormatNormalizeOptions(requestedInternalFormat,
+                                                            GetDriverPixelFormatNormalizeOptions());
+        }
+
+        Bool HasCachedFormatCapability(TextureInternalFormat internalFormat,
+                                       SizeT targetIndex,
+                                       Bool caveat,
+                                       FormatCapability capability) {
+            if (!pActiveBackendObject || targetIndex >= kFormatCapabilityTargetCount) {
+                return false;
+            }
+            const SizeT formatIndex = static_cast<SizeT>(internalFormat);
+            if (formatIndex >= kFormatCapabilityFormatCount) {
+                return false;
+            }
+
+            const FormatCapabilityCache& cache = pActiveBackendObject->GetFormatCapabilities();
+            const FormatCapabilityFlags caps =
+                caveat ? cache.CaveatCaps[targetIndex][formatIndex] : cache.FullCaps[targetIndex][formatIndex];
+            return HasFormatCapability(caps, capability);
+        }
+
+        Bool HasAnyCachedFormatCapability(TextureInternalFormat internalFormat,
+                                          Bool caveat,
+                                          FormatCapability capability) {
+            for (SizeT targetIndex = 0; targetIndex < kFormatCapabilityTargetCount; ++targetIndex) {
+                if (HasCachedFormatCapability(internalFormat, targetIndex, caveat, capability)) {
+                    return true;
+                }
+            }
+            return false;
+        }
+
+        Bool ShouldUseCaveatFormat(TextureInternalFormat internalFormat, SizeT targetIndex) {
+            if (targetIndex < kFormatCapabilityTargetCount) {
+                const Bool fullCreatable =
+                    HasCachedFormatCapability(internalFormat, targetIndex, false, FormatCapability::Creatable);
+                const Bool caveatCreatable =
+                    HasCachedFormatCapability(internalFormat, targetIndex, true, FormatCapability::Creatable);
+                const Bool fullRenderable =
+                    HasCachedFormatCapability(internalFormat, targetIndex, false, FormatCapability::FramebufferRenderable);
+                const Bool caveatRenderable =
+                    HasCachedFormatCapability(internalFormat, targetIndex, true, FormatCapability::FramebufferRenderable);
+                return (!fullCreatable && caveatCreatable) || (!fullRenderable && caveatRenderable);
+            }
+
+            if (HasAnyCachedFormatCapability(internalFormat, false, FormatCapability::Creatable)) {
+                return false;
+            }
+            return HasAnyCachedFormatCapability(internalFormat, true, FormatCapability::Creatable);
+        }
+
+        void GenerateFormatInfo(TextureInternalFormat internalFormat,
+                                SizeT targetIndex,
+                                GLenum* outInternalFormat,
+                                GLenum* outFormat,
+                                GLenum* outType) {
+            using namespace MobileGL::MG_Util::TextureFormatProcessor;
+            const GLenum requestedInternalFormat = MG_Util::ConvertTextureInternalFormatToGLEnum(internalFormat);
+            Flags<PixelFormatNormalizeOptionBit> options;
+            if (!pActiveBackendObject || ShouldUseCaveatFormat(internalFormat, targetIndex)) {
+                options = GetRuntimeFallbackNormalizeOptions(requestedInternalFormat);
+            }
+            NormalizePixelFormat(requestedInternalFormat, options, outInternalFormat, outFormat, outType);
+        }
+    } // namespace
+
     namespace TextureImpl {
         void GenerateTextureFormatInfo(TextureInternalFormat internalFormat, GLenum* outInternalFormat,
-                                       GLenum* outFormat, GLenum* outType) {
+                                       GLenum* outFormat, GLenum* outType, TextureTarget target) {
 #ifdef TRACY_ENABLE
             ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
 #endif
-            using namespace MobileGL::MG_Util::TextureFormatProcessor;
-            auto options = (g_GLESCapabilities.SupportsNorm16Texture) ? PixelFormatNormalizeOptionBit::None
-                                                                      : PixelFormatNormalizeOptionBit::NoNorm16;
-            NormalizePixelFormat(MG_Util::ConvertTextureInternalFormatToGLEnum(internalFormat), options,
-                                 outInternalFormat, outFormat, outType);
+            const SizeT targetIndex =
+                target == TextureTarget::Unknown ? kFormatCapabilityTargetCount : GetFormatCapabilityTargetIndex(target);
+            GenerateFormatInfo(internalFormat, targetIndex, outInternalFormat, outFormat, outType);
+        }
+
+        void GenerateRenderbufferFormatInfo(TextureInternalFormat internalFormat, GLenum* outInternalFormat,
+                                            GLenum* outFormat, GLenum* outType) {
+#ifdef TRACY_ENABLE
+            ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+            GenerateFormatInfo(internalFormat, GetRenderbufferFormatCapabilityTargetIndex(), outInternalFormat,
+                               outFormat, outType);
+        }
+
+        Bool ShouldUseCaveatTextureFormat(TextureInternalFormat internalFormat, TextureTarget target) {
+            const SizeT targetIndex =
+                target == TextureTarget::Unknown ? kFormatCapabilityTargetCount : GetFormatCapabilityTargetIndex(target);
+            return ShouldUseCaveatFormat(internalFormat, targetIndex);
+        }
+
+        Bool ShouldUseCaveatRenderbufferFormat(TextureInternalFormat internalFormat) {
+            return ShouldUseCaveatFormat(internalFormat, GetRenderbufferFormatCapabilityTargetIndex());
         }
     } // namespace TextureImpl
     namespace PrgramImpl {
@@ -101,14 +218,163 @@ namespace MobileGL::MG_Backend::DirectGLES {
             return result;
         }
 
+        String ClampNormFallbackOutputs(String glslCode, GLenum shaderType, Uint32 snormOutputMask,
+                                        Uint32 unormOutputMask) {
+#ifdef TRACY_ENABLE
+            ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+            const Uint32 outputMask = snormOutputMask | unormOutputMask;
+            if (shaderType != GL_FRAGMENT_SHADER || outputMask == 0) {
+                return glslCode;
+            }
+
+            const std::regex outputPattern(
+                R"(layout\s*\(\s*location\s*=\s*([0-9]+)\s*\)\s*out\s+(?:(?:lowp|mediump|highp)\s+)?vec4\s+([A-Za-z_][A-Za-z0-9_]*)\s*;)");
+            std::sregex_iterator outputIt(glslCode.begin(), glslCode.end(), outputPattern);
+            std::sregex_iterator outputEnd;
+            struct OutputClamp {
+                String Name;
+                Bool Signed;
+            };
+            Vector<OutputClamp> outputClamps;
+            for (; outputIt != outputEnd; ++outputIt) {
+                const Uint location = static_cast<Uint>(std::stoul((*outputIt)[1].str()));
+                if (location < 32 && (outputMask & (1u << location))) {
+                    outputClamps.push_back({(*outputIt)[2].str(), static_cast<Bool>(snormOutputMask & (1u << location))});
+                }
+            }
+            if (outputClamps.empty()) {
+                return glslCode;
+            }
+
+            const std::regex mainPattern(R"(void\s+main\s*\([^)]*\)\s*\{)");
+            std::smatch mainMatch;
+            if (!std::regex_search(glslCode, mainMatch, mainPattern)) {
+                return glslCode;
+            }
+
+            SizeT bracePos = static_cast<SizeT>(mainMatch.position(0) + mainMatch.length(0) - 1);
+            Int depth = 0;
+            for (SizeT pos = bracePos; pos < glslCode.size(); ++pos) {
+                if (glslCode[pos] == '{') {
+                    ++depth;
+                } else if (glslCode[pos] == '}') {
+                    --depth;
+                    if (depth == 0) {
+                        String clampLine;
+                        for (const OutputClamp& outputClamp : outputClamps) {
+                            const String minValue = outputClamp.Signed ? "-1.0" : "0.0";
+                            clampLine += "\n    " + outputClamp.Name + " = clamp(" + outputClamp.Name +
+                                         ", vec4(" + minValue + "), vec4(1.0));";
+                        }
+                        clampLine += "\n";
+                        glslCode.insert(pos, clampLine);
+                        return glslCode;
+                    }
+                }
+            }
+            return glslCode;
+        }
+
+        String ForceFlatIntegerVaryings(const String& glslCode, GLenum shaderType) {
+#ifdef TRACY_ENABLE
+            ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+            String result = glslCode;
+            const String integerType = R"((?:(?:lowp|mediump|highp)\s+)?(?:u?int|[iu]vec[234])\b)";
+
+            auto addFlatQualifier = [&result, &integerType](const String& qualifier) {
+                const std::regex pattern("(layout\\s*\\([^)]*\\)\\s*)(?!(?:flat|smooth|noperspective)\\s)(" +
+                                         qualifier + "\\s+" + integerType + ")");
+                result = std::regex_replace(result, pattern, "$1flat $2");
+            };
+
+            switch (shaderType) {
+            case GL_VERTEX_SHADER:
+                addFlatQualifier("out");
+                break;
+            case GL_GEOMETRY_SHADER:
+                addFlatQualifier("in");
+                addFlatQualifier("out");
+                break;
+            case GL_FRAGMENT_SHADER:
+                addFlatQualifier("in");
+                break;
+            default:
+                break;
+            }
+
+            return result;
+        }
+
         String RemoveLayoutBinding(const String& glslCode) {
 #ifdef TRACY_ENABLE
             ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
 #endif
+            // Sampler and uniform-block bindings are re-established at draw time through the
+            // API, so their layout qualifiers are stripped (they may exceed ES limits). SSBO
+            // blocks and image uniforms are different: ES has no glShaderStorageBlockBinding,
+            // and image units cannot be set with glUniform1i, so for those declarations the
+            // binding qualifier is the only binding mechanism and must be preserved.
             static std::regex bindingRegex(R"(layout\s*\(\s*binding\s*=\s*\d+\s*\)\s*)");
-            String result = std::regex_replace(glslCode, bindingRegex, "");
             static std::regex bindingRegex2(R"(layout\s*\(\s*binding\s*=\s*\d+\s*,)");
-            result = std::regex_replace(result, bindingRegex2, "layout(");
+            static std::regex keepBindingRegex(R"(\b(buffer|[iu]?image[A-Za-z0-9]*)\b)");
+
+            String result;
+            result.reserve(glslCode.size());
+            SizeT lineStart = 0;
+            while (lineStart <= glslCode.size()) {
+                SizeT lineEnd = glslCode.find('\n', lineStart);
+                const Bool lastLine = lineEnd == String::npos;
+                String line = glslCode.substr(lineStart, lastLine ? String::npos : lineEnd - lineStart);
+
+                if (!std::regex_search(line, keepBindingRegex)) {
+                    line = std::regex_replace(line, bindingRegex, "");
+                    line = std::regex_replace(line, bindingRegex2, "layout(");
+                }
+
+                result += line;
+                if (lastLine) {
+                    break;
+                }
+                result += '\n';
+                lineStart = lineEnd + 1;
+            }
+            return result;
+        }
+
+        String RemoveClipDistanceRedeclaration(const String& glslCode) {
+#ifdef TRACY_ENABLE
+            ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+            // Adreno rejects any redeclaration of gl_ClipDistance/gl_CullDistance ("reserved
+            // built-in name") even with GL_EXT_clip_cull_distance required, but accepts plain
+            // usage of the builtin. Drop the desktop-style redeclaration line SPIRV-Cross
+            // prints; the "#extension GL_EXT_clip_cull_distance : require" line stays.
+            static const std::regex redeclarationRegex(
+                R"(^\s*(?:out|in)\s+(?:(?:high|medium|low)p\s+)?float\s+gl_(?:Clip|Cull)Distance\[[0-9]+\];\s*$)");
+
+            String result;
+            result.reserve(glslCode.size());
+            SizeT lineStart = 0;
+            Bool firstLine = true;
+            while (lineStart <= glslCode.size()) {
+                SizeT lineEnd = glslCode.find('\n', lineStart);
+                const Bool lastLine = lineEnd == String::npos;
+                String line = glslCode.substr(lineStart, lastLine ? String::npos : lineEnd - lineStart);
+
+                if (!std::regex_match(line, redeclarationRegex)) {
+                    if (!firstLine) {
+                        result += '\n';
+                    }
+                    result += line;
+                    firstLine = false;
+                }
+                if (lastLine) {
+                    break;
+                }
+                lineStart = lineEnd + 1;
+            }
             return result;
         }
     } // namespace PrgramImpl
@@ -118,7 +384,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
 #ifdef TRACY_ENABLE
             ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
 #endif
-            while (GLenum err = g_GLESFuncs.glGetError() != GL_NO_ERROR) {
+            for (GLenum err = g_GLESFuncs.glGetError(); err != GL_NO_ERROR; err = g_GLESFuncs.glGetError()) {
                 MGLOG_E("-> GLES Error: %s", MG_Util::ConvertGLEnumToString(err).c_str());
             }
         }
@@ -221,4 +487,407 @@ namespace MobileGL::MG_Backend::DirectGLES {
             }
         }
     } // namespace Utils
+
+    // ---- Client-format readback conversion helpers -------------------------------------------------
+    // ReadPixels/GetTexImage read a guaranteed wide RGBA(_INTEGER) layout from the ES driver and repack
+    // it on the CPU into the client's (format, type) layout. Everything here is pure byte shuffling so
+    // unit tests can assert the exact packed words; field positions follow GL 3.3 table 3.6 and mirror
+    // the GL CTS packed_pixels oracle (glcPackedPixelsTests.cpp pack_UNSIGNED_* helpers).
+    namespace ReadbackImpl {
+        using MG_Util::DecodeHalfBitsToFloat;
+        using MG_Util::EncodeFloatToHalfBits;
+
+        Bool GetReadbackChannelMapping(GLenum format, ReadbackChannelMapping& outMapping) {
+            switch (format) {
+            case GL_RED:          outMapping = {{0, 0, 0, 0}, 1, false}; return true;
+            case GL_RED_INTEGER:  outMapping = {{0, 0, 0, 0}, 1, true};  return true;
+            // Desktop-GL single-channel client formats (GL CTS packed_pixels rgba8_format_green/blue):
+            // the destination holds one component sourced from the named channel of the wide RGBA read.
+            // GL_ALPHA is mapped here from the raw enum because the state layer folds it into Red for the
+            // legacy alpha-texture upload hack.
+            case GL_GREEN:         outMapping = {{1, 0, 0, 0}, 1, false}; return true;
+            case GL_GREEN_INTEGER: outMapping = {{1, 0, 0, 0}, 1, true};  return true;
+            case GL_BLUE:          outMapping = {{2, 0, 0, 0}, 1, false}; return true;
+            case GL_BLUE_INTEGER:  outMapping = {{2, 0, 0, 0}, 1, true};  return true;
+            case GL_ALPHA:         outMapping = {{3, 0, 0, 0}, 1, false}; return true;
+            case GL_ALPHA_INTEGER: outMapping = {{3, 0, 0, 0}, 1, true};  return true;
+            case GL_RG:           outMapping = {{0, 1, 0, 0}, 2, false}; return true;
+            case GL_RG_INTEGER:   outMapping = {{0, 1, 0, 0}, 2, true};  return true;
+            case GL_RGB:          outMapping = {{0, 1, 2, 0}, 3, false}; return true;
+            case GL_RGB_INTEGER:  outMapping = {{0, 1, 2, 0}, 3, true};  return true;
+            case GL_BGR:          outMapping = {{2, 1, 0, 0}, 3, false}; return true;
+            case GL_BGR_INTEGER:  outMapping = {{2, 1, 0, 0}, 3, true};  return true;
+            case GL_RGBA:         outMapping = {{0, 1, 2, 3}, 4, false}; return true;
+            case GL_RGBA_INTEGER: outMapping = {{0, 1, 2, 3}, 4, true};  return true;
+            case GL_BGRA:         outMapping = {{2, 1, 0, 3}, 4, false}; return true;
+            case GL_BGRA_INTEGER: outMapping = {{2, 1, 0, 3}, 4, true};  return true;
+            default:
+                return false;
+            }
+        }
+
+        Bool GetPackedReadbackLayout(GLenum type, PackedReadbackLayout& out) {
+            switch (type) {
+            // Non-REV types pack the first format component starting at the most significant bit,
+            // *_REV types starting at the least significant bit (GL CTS pack_UNSIGNED_SHORT_5_6_5:
+            // R bits 15-11; pack_UNSIGNED_SHORT_1_5_5_5_REV: R bits 4-0, A bit 15).
+            case GL_UNSIGNED_BYTE_3_3_2:          out = {3, {3, 3, 2, 0},    {5, 2, 0, 0},    1, false}; return true;
+            case GL_UNSIGNED_BYTE_2_3_3_REV:      out = {3, {3, 3, 2, 0},    {0, 3, 6, 0},    1, false}; return true;
+            case GL_UNSIGNED_SHORT_5_6_5:         out = {3, {5, 6, 5, 0},    {11, 5, 0, 0},   2, false}; return true;
+            case GL_UNSIGNED_SHORT_5_6_5_REV:     out = {3, {5, 6, 5, 0},    {0, 5, 11, 0},   2, false}; return true;
+            case GL_UNSIGNED_SHORT_4_4_4_4:       out = {4, {4, 4, 4, 4},    {12, 8, 4, 0},   2, false}; return true;
+            case GL_UNSIGNED_SHORT_4_4_4_4_REV:   out = {4, {4, 4, 4, 4},    {0, 4, 8, 12},   2, false}; return true;
+            case GL_UNSIGNED_SHORT_5_5_5_1:       out = {4, {5, 5, 5, 1},    {11, 6, 1, 0},   2, false}; return true;
+            case GL_UNSIGNED_SHORT_1_5_5_5_REV:   out = {4, {5, 5, 5, 1},    {0, 5, 10, 15},  2, false}; return true;
+            case GL_UNSIGNED_INT_8_8_8_8:         out = {4, {8, 8, 8, 8},    {24, 16, 8, 0},  4, false}; return true;
+            case GL_UNSIGNED_INT_8_8_8_8_REV:     out = {4, {8, 8, 8, 8},    {0, 8, 16, 24},  4, false}; return true;
+            case GL_UNSIGNED_INT_10_10_10_2:      out = {4, {10, 10, 10, 2}, {22, 12, 2, 0},  4, false}; return true;
+            case GL_UNSIGNED_INT_2_10_10_10_REV:  out = {4, {10, 10, 10, 2}, {0, 10, 20, 30}, 4, false}; return true;
+            // Packed-float RGB types: fields hold unsigned small floats; 5_9_9_9_REV's shared 5-bit
+            // exponent (bits 31-27) is emitted by EncodeSharedExponentRGB9E5, not a component field.
+            case GL_UNSIGNED_INT_10F_11F_11F_REV: out = {3, {11, 11, 10, 0}, {0, 11, 22, 0},  4, true};  return true;
+            case GL_UNSIGNED_INT_5_9_9_9_REV:     out = {3, {9, 9, 9, 0},    {0, 9, 18, 0},   4, true};  return true;
+            default:
+                return false;
+            }
+        }
+
+        SizeT GetReadbackComponentSize(GLenum type) {
+            PackedReadbackLayout packedLayout{};
+            if (GetPackedReadbackLayout(type, packedLayout)) {
+                return packedLayout.byteSize;
+            }
+            switch (type) {
+            case GL_UNSIGNED_BYTE:
+            case GL_BYTE:
+                return 1;
+            case GL_UNSIGNED_SHORT:
+            case GL_SHORT:
+            case GL_HALF_FLOAT:
+                return 2;
+            case GL_UNSIGNED_INT:
+            case GL_INT:
+            case GL_FLOAT:
+                return 4;
+            default:
+                return 0;
+            }
+        }
+
+        SizeT GetReadbackDstPixelSize(const ReadbackChannelMapping& mapping, GLenum type) {
+            PackedReadbackLayout packedLayout{};
+            if (GetPackedReadbackLayout(type, packedLayout)) {
+                if (packedLayout.fieldCount != mapping.channelCount) {
+                    return 0; // 3-field packed types pair with 3-component formats only, 4 with 4
+                }
+                if (mapping.isInteger && packedLayout.isFloatPacked) {
+                    return 0; // packed-float RGB types never pair with integer formats
+                }
+                return packedLayout.byteSize;
+            }
+            if (mapping.isInteger && (type == GL_FLOAT || type == GL_HALF_FLOAT)) {
+                return 0;
+            }
+            const SizeT componentSize = GetReadbackComponentSize(type);
+            return componentSize == 0 ? 0 : static_cast<SizeT>(mapping.channelCount) * componentSize;
+        }
+
+        namespace {
+            void WritePackedReadbackWord(Uint8* dst, Uint32 word, SizeT byteSize) {
+                switch (byteSize) {
+                case 1: {
+                    const auto out = static_cast<Uint8>(word);
+                    Memcpy(dst, &out, sizeof(out));
+                    break;
+                }
+                case 2: {
+                    const auto out = static_cast<Uint16>(word);
+                    Memcpy(dst, &out, sizeof(out));
+                    break;
+                }
+                default:
+                    Memcpy(dst, &word, sizeof(word));
+                    break;
+                }
+            }
+        } // namespace
+
+        // Shared encoders live in MG_Util/Math/SmallFloat.h so the upload conversion
+        // (PixelStoreProcessor) uses byte-identical packing; kept exported here for unit tests.
+        Uint32 EncodeFloatToUnsignedF11(Float value) { return MG_Util::EncodeFloatToUnsignedF11(value); }
+        Uint32 EncodeFloatToUnsignedF10(Float value) { return MG_Util::EncodeFloatToUnsignedF10(value); }
+        Uint32 EncodeSharedExponentRGB9E5(const Float rgb[3]) { return MG_Util::EncodeSharedExponentRGB9E5(rgb); }
+
+        void ConvertWideReadbackRow(const Uint8* src, Uint8* dst, SizeT width, GLenum wideType,
+                                    const ReadbackChannelMapping& mapping, GLenum type) {
+            PackedReadbackLayout packedLayout{};
+            const Bool isPacked = GetPackedReadbackLayout(type, packedLayout);
+            const SizeT dstComponentSize = GetReadbackComponentSize(type);
+            const SizeT dstPixelBytes = GetReadbackDstPixelSize(mapping, type);
+            const SizeT srcPixelBytes = 4 * GetReadbackComponentSize(wideType);
+
+            for (SizeT col = 0; col < width; ++col) {
+                const Uint8* srcPixel = src + col * srcPixelBytes;
+                Uint8* dstPixel = dst + col * dstPixelBytes;
+                if (mapping.isInteger) {
+                    Int64 srcValues[4];
+                    for (Int c = 0; c < 4; ++c) {
+                        srcValues[c] = wideType == GL_INT
+                                           ? static_cast<Int64>(reinterpret_cast<const Int32*>(srcPixel)[c])
+                                           : static_cast<Int64>(reinterpret_cast<const Uint32*>(srcPixel)[c]);
+                    }
+                    if (isPacked) {
+                        // Integer sources clamp each component to the unsigned range of its field
+                        // (GL 3.3 section 4.3.1 final conversion).
+                        Uint32 word = 0;
+                        for (Int ch = 0; ch < packedLayout.fieldCount; ++ch) {
+                            const Int64 fieldMax = (Int64{1} << packedLayout.width[ch]) - 1;
+                            const auto v = static_cast<Uint32>(
+                                std::clamp<Int64>(srcValues[mapping.sourceChannel[ch]], 0, fieldMax));
+                            word |= v << packedLayout.shift[ch];
+                        }
+                        WritePackedReadbackWord(dstPixel, word, packedLayout.byteSize);
+                    } else {
+                        for (Int ch = 0; ch < mapping.channelCount; ++ch) {
+                            const Int64 v = srcValues[mapping.sourceChannel[ch]];
+                            Uint8* dstComponent = dstPixel + static_cast<SizeT>(ch) * dstComponentSize;
+                            switch (type) {
+                            case GL_UNSIGNED_BYTE:
+                                *dstComponent = static_cast<Uint8>(std::clamp<Int64>(v, 0, 255));
+                                break;
+                            case GL_BYTE: {
+                                const auto out = static_cast<Int8>(std::clamp<Int64>(v, -128, 127));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_UNSIGNED_SHORT: {
+                                const auto out = static_cast<Uint16>(std::clamp<Int64>(v, 0, 65535));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_SHORT: {
+                                const auto out = static_cast<Int16>(std::clamp<Int64>(v, -32768, 32767));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_UNSIGNED_INT: {
+                                const auto out = static_cast<Uint32>(std::clamp<Int64>(v, 0, 4294967295LL));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_INT: {
+                                const auto out =
+                                    static_cast<Int32>(std::clamp<Int64>(v, -2147483648LL, 2147483647LL));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            default:
+                                break;
+                            }
+                        }
+                    }
+                } else {
+                    Float srcValues[4];
+                    switch (wideType) {
+                    case GL_UNSIGNED_BYTE:
+                        for (Int c = 0; c < 4; ++c) {
+                            srcValues[c] = static_cast<Float>(srcPixel[c]) / 255.0f;
+                        }
+                        break;
+                    case GL_BYTE:
+                        for (Int c = 0; c < 4; ++c) {
+                            srcValues[c] = std::max(
+                                static_cast<Float>(reinterpret_cast<const Int8*>(srcPixel)[c]) / 127.0f, -1.0f);
+                        }
+                        break;
+                    case GL_UNSIGNED_SHORT:
+                        for (Int c = 0; c < 4; ++c) {
+                            srcValues[c] =
+                                static_cast<Float>(reinterpret_cast<const Uint16*>(srcPixel)[c]) / 65535.0f;
+                        }
+                        break;
+                    case GL_SHORT:
+                        for (Int c = 0; c < 4; ++c) {
+                            srcValues[c] = std::max(
+                                static_cast<Float>(reinterpret_cast<const Int16*>(srcPixel)[c]) / 32767.0f, -1.0f);
+                        }
+                        break;
+                    case GL_HALF_FLOAT:
+                        for (Int c = 0; c < 4; ++c) {
+                            srcValues[c] = DecodeHalfBitsToFloat(reinterpret_cast<const Uint16*>(srcPixel)[c]);
+                        }
+                        break;
+                    default: // GL_FLOAT
+                        for (Int c = 0; c < 4; ++c) {
+                            srcValues[c] = reinterpret_cast<const Float*>(srcPixel)[c];
+                        }
+                        break;
+                    }
+                    if (isPacked) {
+                        Uint32 word = 0;
+                        if (packedLayout.isFloatPacked) {
+                            const Float fields[3] = {srcValues[mapping.sourceChannel[0]],
+                                                     srcValues[mapping.sourceChannel[1]],
+                                                     srcValues[mapping.sourceChannel[2]]};
+                            word = type == GL_UNSIGNED_INT_5_9_9_9_REV
+                                       ? EncodeSharedExponentRGB9E5(fields)
+                                       : (EncodeFloatToUnsignedF11(fields[0]) << packedLayout.shift[0]) |
+                                             (EncodeFloatToUnsignedF11(fields[1]) << packedLayout.shift[1]) |
+                                             (EncodeFloatToUnsignedF10(fields[2]) << packedLayout.shift[2]);
+                        } else {
+                            // Normalized encode: round(clamp(v, 0, 1) * (2^bits - 1)) into each field.
+                            for (Int ch = 0; ch < packedLayout.fieldCount; ++ch) {
+                                const auto fieldMax = static_cast<Float>((1u << packedLayout.width[ch]) - 1u);
+                                const auto v = static_cast<Uint32>(std::llround(
+                                    std::clamp(srcValues[mapping.sourceChannel[ch]], 0.0f, 1.0f) * fieldMax));
+                                word |= v << packedLayout.shift[ch];
+                            }
+                        }
+                        WritePackedReadbackWord(dstPixel, word, packedLayout.byteSize);
+                    } else {
+                        for (Int ch = 0; ch < mapping.channelCount; ++ch) {
+                            const Float v = srcValues[mapping.sourceChannel[ch]];
+                            Uint8* dstComponent = dstPixel + static_cast<SizeT>(ch) * dstComponentSize;
+                            switch (type) {
+                            case GL_UNSIGNED_BYTE:
+                                *dstComponent =
+                                    static_cast<Uint8>(std::llround(std::clamp(v, 0.0f, 1.0f) * 255.0));
+                                break;
+                            case GL_BYTE: {
+                                const auto out =
+                                    static_cast<Int8>(std::llround(std::clamp(v, -1.0f, 1.0f) * 127.0));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_UNSIGNED_SHORT: {
+                                const auto out =
+                                    static_cast<Uint16>(std::llround(std::clamp(v, 0.0f, 1.0f) * 65535.0));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_SHORT: {
+                                const auto out =
+                                    static_cast<Int16>(std::llround(std::clamp(v, -1.0f, 1.0f) * 32767.0));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_UNSIGNED_INT: {
+                                const auto out = static_cast<Uint32>(
+                                    std::llround(static_cast<Double>(std::clamp(v, 0.0f, 1.0f)) * 4294967295.0));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_INT: {
+                                const auto out = static_cast<Int32>(
+                                    std::llround(static_cast<Double>(std::clamp(v, -1.0f, 1.0f)) * 2147483647.0));
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            case GL_FLOAT:
+                                Memcpy(dstComponent, &v, sizeof(v));
+                                break;
+                            case GL_HALF_FLOAT: {
+                                const Uint16 out = EncodeFloatToHalfBits(v);
+                                Memcpy(dstComponent, &out, sizeof(out));
+                                break;
+                            }
+                            default:
+                                break;
+                            }
+                        }
+                    }
+                }
+            }
+        }
+
+        static SizeT AlignReadbackRow(SizeT rowBytes, Int alignment) {
+            const SizeT align = alignment > 0 ? static_cast<SizeT>(alignment) : 1;
+            return (rowBytes + align - 1) / align * align;
+        }
+
+    // Repacks wide RGBA(_INTEGER) rows into the client's (format, type) layout, honoring the
+        // client-side PACK parameters and the bound pixel-pack buffer. `wide` holds
+        // `sliceHeight * sliceCount` rows of `width` texels (slice-major, tightly stacked),
+        // 4 components x GetReadbackComponentSize(wideType) bytes each.
+        // applyPackImageParams: GL_PACK_IMAGE_HEIGHT / GL_PACK_SKIP_IMAGES apply only to GetTexImage
+        // of 3D/array images; ReadPixels and 2D GetTexImage ignore them (GL 3.3 sections 4.3.1, 6.1.4).
+        // Per the GL addressing rules, slice k row j lands at
+        // SKIP_IMAGES*imageStride + SKIP_ROWS*rowStride + SKIP_PIXELS*pixelBytes
+        //   + k*imageStride + j*rowStride, with imageStride = max(IMAGE_HEIGHT, sliceHeight)*rowStride.
+        Bool StoreWideRowsToClient(const Uint8* wide, GLenum wideType, GLsizei width, GLsizei sliceHeight,
+                                   GLsizei sliceCount, const ReadbackChannelMapping& mapping, GLenum type,
+                                   void* pixels, Bool applyPackImageParams) {
+        const SizeT dstPixelBytes = GetReadbackDstPixelSize(mapping, type);
+        if (dstPixelBytes == 0) {
+            return false;
+        }
+        PackedReadbackLayout packedLayout{};
+        const Bool isPackedType = GetPackedReadbackLayout(type, packedLayout);
+        const SizeT dstComponentSize = GetReadbackComponentSize(type);
+
+        const auto& pixelPackBufferObject =
+            MG_State::pGLContext->GetBufferBindingSlot(BufferTarget::PixelPack).GetBoundObject();
+
+        // Destination layout is computed from the client-side PACK parameters; only the actual pixel
+        // rows are written so skip regions of the destination stay untouched.
+        const auto packParams = MG_State::pGLContext->GetPixelStoreParameters(false);
+        const SizeT rowPixels = static_cast<SizeT>(packParams.RowLength > 0 ? packParams.RowLength : width);
+        const SizeT dstRowStride = AlignReadbackRow(rowPixels * dstPixelBytes, packParams.Alignment);
+        const SizeT imageRows =
+            applyPackImageParams && packParams.ImageHeight > 0
+                ? static_cast<SizeT>(packParams.ImageHeight)
+                : static_cast<SizeT>(sliceHeight);
+        const SizeT dstImageStride = imageRows * dstRowStride;
+        const SizeT skipImages =
+            applyPackImageParams ? static_cast<SizeT>(std::max(packParams.SkipImages, 0)) : SizeT{0};
+        const SizeT dstSkipOffset = skipImages * dstImageStride +
+                                    static_cast<SizeT>(std::max(packParams.SkipRows, 0)) * dstRowStride +
+                                    static_cast<SizeT>(std::max(packParams.SkipPixels, 0)) * dstPixelBytes;
+        const SizeT dstRowBytes = static_cast<SizeT>(width) * dstPixelBytes;
+
+        const SizeT pboBaseOffset = reinterpret_cast<SizeT>(pixels); // with a PBO, `pixels` is an offset
+        if (pixelPackBufferObject) {
+            const SizeT requiredSize = pboBaseOffset + dstSkipOffset +
+                                    static_cast<SizeT>(sliceCount - 1) * dstImageStride +
+                                    static_cast<SizeT>(sliceHeight - 1) * dstRowStride + dstRowBytes;
+            if (requiredSize > pixelPackBufferObject->GetSize()) {
+                MGLOG_E("Readback conversion: pixel pack buffer is too small");
+                return true;
+            }
+        }
+
+        const SizeT srcComponentSize = GetReadbackComponentSize(wideType);
+        const SizeT srcPixelBytes = 4 * srcComponentSize;
+        Vector<Uint8> convertedRow(dstRowBytes);
+
+        for (GLsizei slice = 0; slice < sliceCount; ++slice) {
+            for (GLsizei row = 0; row < sliceHeight; ++row) {
+                const SizeT flatRow = static_cast<SizeT>(slice) * static_cast<SizeT>(sliceHeight) +
+                                   static_cast<SizeT>(row);
+                const Uint8* srcRow = wide + flatRow * static_cast<SizeT>(width) * srcPixelBytes;
+                ConvertWideReadbackRow(srcRow, convertedRow.data(), static_cast<SizeT>(width), wideType,
+                                                  mapping, type);
+
+                if (packParams.SwapBytes) {
+                    const SizeT groupSize = isPackedType ? packedLayout.byteSize : dstComponentSize;
+                    if (groupSize > 1) {
+                        for (SizeT offset = 0; offset + groupSize <= dstRowBytes; offset += groupSize) {
+                            std::reverse(convertedRow.data() + offset, convertedRow.data() + offset + groupSize);
+                        }
+                    }
+                }
+
+                const SizeT dstOffset = dstSkipOffset + static_cast<SizeT>(slice) * dstImageStride +
+                                     static_cast<SizeT>(row) * dstRowStride;
+                if (pixelPackBufferObject) {
+                    pixelPackBufferObject->WritebackFromBackend({convertedRow.data(), dstRowBytes},
+                                                             pboBaseOffset + dstOffset);
+                } else {
+                    Memcpy(static_cast<Uint8*>(pixels) + dstOffset, convertedRow.data(), dstRowBytes);
+                }
+            }
+        }
+            return true;
+        }
+    } // namespace ReadbackImpl
 } // namespace MobileGL::MG_Backend::DirectGLES

MobileGL/MG_Backend/DirectGLES/Utils.h

@@ -35,15 +35,77 @@ namespace MobileGL::MG_Backend::DirectGLES {
 
     namespace TextureImpl {
         void GenerateTextureFormatInfo(TextureInternalFormat internalFormat, GLenum* outInternalFormat,
-                                       GLenum* outFormat, GLenum* outType);
+                                       GLenum* outFormat, GLenum* outType,
+                                       TextureTarget target = TextureTarget::Unknown);
+        void GenerateRenderbufferFormatInfo(TextureInternalFormat internalFormat, GLenum* outInternalFormat,
+                                            GLenum* outFormat, GLenum* outType);
+        Bool ShouldUseCaveatTextureFormat(TextureInternalFormat internalFormat, TextureTarget target);
+        Bool ShouldUseCaveatRenderbufferFormat(TextureInternalFormat internalFormat);
     } // namespace TextureImpl
 
     namespace FramebufferImpl {} // namespace FramebufferImpl
 
+    // Pure CPU helpers of the client-format readback conversion (ReadPixels/GetTexImage repack a wide
+    // RGBA(_INTEGER) read into the caller's (format, type) layout). Kept context-free so unit tests can
+    // exercise the exact packing the GL CTS packed_pixels oracle compares against.
+    namespace ReadbackImpl {
+        struct ReadbackChannelMapping {
+            Int sourceChannel[4]; // RGBA source channel feeding each destination component
+            Int channelCount;     // destination component count
+            Bool isInteger;
+        };
+        Bool GetReadbackChannelMapping(GLenum format, ReadbackChannelMapping& outMapping);
+
+        // Byte size of one destination component of `type`; packed types report the packed word size.
+        // 0 = type not supported by the conversion path.
+        SizeT GetReadbackComponentSize(GLenum type);
+
+        // Bit-field layout of a GL packed pixel type. width/shift are indexed in the client format's
+        // component order (matching ReadbackChannelMapping); shift is the LSB position of the field in
+        // the packed word: non-REV types pack the first component from the MSB, *_REV types from the
+        // LSB (GL 3.3 table 3.6; field positions mirror the GL CTS glcPackedPixelsTests pack_* oracle).
+        struct PackedReadbackLayout {
+            Int fieldCount;     // format components stored in the packed word
+            Int width[4];       // bit width of each component's field
+            Int shift[4];       // LSB bit position of each component's field
+            SizeT byteSize;     // packed word size in bytes (1, 2 or 4)
+            Bool isFloatPacked; // 10F_11F_11F_REV / 5_9_9_9_REV: fields hold unsigned small floats
+        };
+        Bool GetPackedReadbackLayout(GLenum type, PackedReadbackLayout& out);
+
+        // Unsigned small-float encoders (EXT_packed_float / EXT_texture_shared_exponent semantics).
+        Uint32 EncodeFloatToUnsignedF11(Float value);
+        Uint32 EncodeFloatToUnsignedF10(Float value);
+        Uint32 EncodeSharedExponentRGB9E5(const Float rgb[3]);
+
+        // Destination bytes per pixel for a (format mapping, type) readback pair; 0 when the pair is
+        // not convertible (unknown type, packed field count != format component count, floating-point
+        // or packed-float type with an integer format).
+        SizeT GetReadbackDstPixelSize(const ReadbackChannelMapping& mapping, GLenum type);
+
+        // Repacks one row of wide RGBA(_INTEGER) texels (4 components of wideType each) into the
+        // client's (format, type) layout. src holds width * 4 * GetReadbackComponentSize(wideType)
+        // 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);
+    } // namespace ReadbackImpl
+
     namespace PrgramImpl {
         String ProcessOutColorLocations(const String& glslCode);
         String ForceSupporterOutput(const String& glslCode);
+        String ClampNormFallbackOutputs(String glslCode, GLenum shaderType, Uint32 snormOutputMask,
+                                        Uint32 unormOutputMask);
+        String ForceFlatIntegerVaryings(const String& glslCode, GLenum shaderType);
         String RemoveLayoutBinding(const String& glslCode);
+        String RemoveClipDistanceRedeclaration(const String& glslCode);
     } // namespace PrgramImpl
 
     namespace Utils {

MobileGL/MG_Backend/DirectVulkan/BackendObject_DirectVulkan.cpp

@@ -9,16 +9,334 @@
 #include "BackendObject_DirectVulkan.h"
 #include "MG_Backend/BackendObject.h"
 #include "DirectVulkan.h"
+#include "MG_State/GLState/FramebufferState/FramebufferObject.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 <Config.h>
+#include <cstdlib>
+#include <cstring>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
     namespace {
+        Bool IsR11G11B10FFallbackEnabled() {
+            return MG_Config::Features.MagmaR11G11B10FFallback;
+        }
+
         Bool IsReleaseCurrentRequest(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx) {
-            return dpy == EGL_NO_DISPLAY && draw == EGL_NO_SURFACE && read == EGL_NO_SURFACE && ctx == EGL_NO_CONTEXT;
+            (void)dpy;
+            return draw == EGL_NO_SURFACE && read == EGL_NO_SURFACE && ctx == EGL_NO_CONTEXT;
+        }
+
+        Bool IsFormatIndexValid(TextureInternalFormat format) {
+            return format != TextureInternalFormat::Unknown && static_cast<Int>(format) >= 0 &&
+                   static_cast<SizeT>(format) < kFormatCapabilityFormatCount;
+        }
+
+        Bool IsLayeredTarget(TextureTarget target) {
+            return target == TextureTarget::Texture3D || target == TextureTarget::Texture1DArray ||
+                   target == TextureTarget::Texture2DArray || target == TextureTarget::TextureCubeMap ||
+                   target == TextureTarget::TextureCubeMapArray ||
+                   target == TextureTarget::Texture2DMultisampleArray;
+        }
+
+        Bool IsMultisampleTarget(TextureTarget target) {
+            return target == TextureTarget::Texture2DMultisample ||
+                   target == TextureTarget::Texture2DMultisampleArray;
+        }
+
+        Bool IsTextureBufferTarget(TextureTarget target) {
+            return target == TextureTarget::TextureBuffer;
+        }
+
+        Bool IsIntegerInternalFormat(TextureInternalFormat format) {
+            const GLenum glFormat = MG_Util::ConvertTextureInternalFormatToGLEnum(format);
+            GLenum normalizedInternalFormat = glFormat;
+            GLenum imageFormat = GL_RGBA;
+            GLenum imageType = GL_UNSIGNED_BYTE;
+            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;
+        }
+
+        FormatCapabilityFlags GetAttachmentCaps(TextureInternalFormat format) {
+            FormatCapabilityFlags caps = FormatCapability::FramebufferRenderable;
+            const Bool isDepth = MG_Util::IsDepthFormatInternalFormat(format);
+            const Bool isStencil = MG_Util::IsStencilFormatInternalFormat(format);
+            if (!isDepth && !isStencil) {
+                caps |= FormatCapability::ColorAttachment;
+            }
+            if (isDepth) {
+                caps |= FormatCapability::DepthAttachment;
+            }
+            if (isStencil) {
+                caps |= FormatCapability::StencilAttachment;
+            }
+            return caps;
+        }
+
+        FormatCapabilityFlags BuildVulkanCaps(TextureInternalFormat logicalFormat,
+                                              TextureTarget target,
+                                              VkFormatFeatureFlags features) {
+            FormatCapabilityFlags caps;
+            const Bool isDepth = MG_Util::IsDepthFormatInternalFormat(logicalFormat);
+            const Bool isStencil = MG_Util::IsStencilFormatInternalFormat(logicalFormat);
+            const Bool isInteger = IsIntegerInternalFormat(logicalFormat);
+
+            if (IsTextureBufferTarget(target)) {
+                if ((features & VK_FORMAT_FEATURE_UNIFORM_TEXEL_BUFFER_BIT) != 0) {
+                    caps |= FormatCapability::Creatable;
+                    caps |= FormatCapability::Sampled;
+                    caps |= FormatCapability::TextureBuffer;
+                }
+                return caps;
+            }
+
+            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 renderable = (isDepth || isStencil) ? depthStencilRenderable : colorRenderable;
+
+            if (sampled || renderable) {
+                caps |= FormatCapability::Creatable;
+            }
+            if (sampled) {
+                caps |= FormatCapability::Sampled;
+                if (linearFilter && !isInteger && !isStencil) {
+                    caps |= FormatCapability::LinearFilter;
+                }
+                if (!isStencil && (features & VK_FORMAT_FEATURE_BLIT_SRC_BIT) != 0 &&
+                    (features & VK_FORMAT_FEATURE_BLIT_DST_BIT) != 0) {
+                    caps |= FormatCapability::GenerateMipmap;
+                }
+                if (!isInteger && !isDepth && !isStencil) {
+                    caps |= FormatCapability::TextureGather;
+                }
+                if (isDepth && !isStencil) {
+                    caps |= FormatCapability::TextureShadow;
+                }
+            }
+            if (renderable) {
+                caps |= GetAttachmentCaps(logicalFormat);
+                if (IsLayeredTarget(target)) {
+                    caps |= FormatCapability::FramebufferLayered;
+                }
+            }
+            if (IsMultisampleTarget(target)) {
+                caps |= FormatCapability::MultisampleTexture;
+            }
+            return caps;
+        }
+
+        Optional<TextureInternalFormat> ResolveVulkanFallbackLogicalFormat(TextureInternalFormat format) {
+            switch (format) {
+            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:
+                return TextureInternalFormat::RGBA8Snorm;
+            case TextureInternalFormat::RGB16:
+                return TextureInternalFormat::RGBA16;
+            case TextureInternalFormat::RGB16Snorm:
+                return TextureInternalFormat::RGBA16Snorm;
+            case TextureInternalFormat::RGB16F:
+                return TextureInternalFormat::RGBA16F;
+            case TextureInternalFormat::R11FG11FB10F:
+                if (IsR11G11B10FFallbackEnabled()) {
+                    return TextureInternalFormat::RGBA16F;
+                }
+                return Nullopt;
+            case TextureInternalFormat::RGB32F:
+                return TextureInternalFormat::RGBA32F;
+            case TextureInternalFormat::RGB8I:
+                return TextureInternalFormat::RGBA8I;
+            case TextureInternalFormat::RGB8UI:
+                return TextureInternalFormat::RGBA8UI;
+            case TextureInternalFormat::RGB16I:
+                return TextureInternalFormat::RGBA16I;
+            case TextureInternalFormat::RGB16UI:
+                return TextureInternalFormat::RGBA16UI;
+            case TextureInternalFormat::RGB32I:
+                return TextureInternalFormat::RGBA32I;
+            case TextureInternalFormat::RGB32UI:
+                return TextureInternalFormat::RGBA32UI;
+            default:
+                return Nullopt;
+            }
+        }
+
+        Optional<VkFormat> ResolveVulkanFallbackFormat(TextureInternalFormat format) {
+            const Optional<TextureInternalFormat> fallbackLogicalFormat = ResolveVulkanFallbackLogicalFormat(format);
+            if (!fallbackLogicalFormat) {
+                return Nullopt;
+            }
+            return MG_Util::ConvertTextureInternalFormatToVkEnum(*fallbackLogicalFormat);
+        }
+
+        Bool HasNewCaveatFormatCaps(FormatCapabilityFlags nativeCaps, FormatCapabilityFlags fallbackCaps) {
+            for (FormatCapability capability : kReportedFormatCapabilities) {
+                if (HasFormatCapability(fallbackCaps, capability) &&
+                    !HasFormatCapability(nativeCaps, capability)) {
+                    return true;
+                }
+            }
+            return false;
+        }
+
+        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());
+        }
+
+        Vector<Int> BuildSampleCounts(Int maxSamples) {
+            Vector<Int> counts;
+            for (Int samples = std::max(maxSamples, 1); samples > 1; samples >>= 1) {
+                counts.push_back(samples);
+            }
+            counts.push_back(1);
+            return counts;
+        }
+
+        void PopulateFormatCapabilitiesImpl(VkPhysicalDevice physicalDevice,
+                                            PFN_vkGetPhysicalDeviceFormatProperties getFormatProperties,
+                                            const MG_External::VulkanCapabilities& capabilities,
+                                            FormatCapabilityCache& cache) {
+            cache.Clear();
+            if (physicalDevice == VK_NULL_HANDLE || getFormatProperties == nullptr) {
+                return;
+            }
+
+            for (SizeT formatIndex = 0; formatIndex < kFormatCapabilityFormatCount; ++formatIndex) {
+                const auto logicalFormat = static_cast<TextureInternalFormat>(formatIndex);
+                if (!IsFormatIndexValid(logicalFormat)) {
+                    continue;
+                }
+
+                VkFormat nativeFormat = MG_Util::ConvertTextureInternalFormatToVkEnum(logicalFormat);
+                const Optional<TextureInternalFormat> fallbackLogicalFormat =
+                    ResolveVulkanFallbackLogicalFormat(logicalFormat);
+                VkFormat fallbackFormat = ResolveVulkanFallbackFormat(logicalFormat).value_or(VK_FORMAT_UNDEFINED);
+
+                VkFormatProperties nativeProperties{};
+                if (nativeFormat != VK_FORMAT_UNDEFINED) {
+                    getFormatProperties(physicalDevice, nativeFormat, &nativeProperties);
+                }
+
+                VkFormatProperties fallbackProperties{};
+                if (fallbackFormat != VK_FORMAT_UNDEFINED && fallbackFormat != nativeFormat) {
+                    getFormatProperties(physicalDevice, fallbackFormat, &fallbackProperties);
+                }
+
+                for (SizeT targetIndex = 0; targetIndex < kFormatCapabilityTextureTargetCount; ++targetIndex) {
+                    const auto target = static_cast<TextureTarget>(targetIndex);
+                    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;
+                    FormatCapabilityFlags fallbackCaps = BuildVulkanCaps(logicalFormat, target, fallbackFeatures);
+                    if (fallbackFormat != VK_FORMAT_UNDEFINED && fallbackFormat != nativeFormat) {
+                        cache.CaveatCaps[targetIndex][formatIndex] |= fallbackCaps;
+                        if (fallbackLogicalFormat && HasNewCaveatFormatCaps(nativeCaps, fallbackCaps)) {
+                            LogVulkanFormatCaveat(logicalFormat, targetIndex, *fallbackLogicalFormat);
+                        }
+                    }
+
+                    if (HasFormatCapability(nativeCaps | fallbackCaps, FormatCapability::MultisampleTexture)) {
+                        const Bool isDepth = MG_Util::IsDepthFormatInternalFormat(logicalFormat);
+                        const Bool isStencil = MG_Util::IsStencilFormatInternalFormat(logicalFormat);
+                        const Bool isInteger = IsIntegerInternalFormat(logicalFormat);
+                        Int maxSamples = capabilities.MaxColorTextureSamples;
+                        if (isDepth || isStencil) {
+                            maxSamples = capabilities.MaxDepthTextureSamples;
+                        } else if (isInteger) {
+                            maxSamples = capabilities.MaxIntegerSamples;
+                        }
+                        cache.SampleCounts[targetIndex][formatIndex] = BuildSampleCounts(maxSamples);
+                    }
+                }
+
+                const SizeT renderbufferTargetIndex = GetRenderbufferFormatCapabilityTargetIndex();
+                FormatCapabilityFlags renderbufferCaps =
+                    BuildVulkanCaps(logicalFormat, TextureTarget::Texture2D, nativeProperties.optimalTilingFeatures);
+                renderbufferCaps &= FormatCapability::Creatable;
+                if ((nativeProperties.optimalTilingFeatures &
+                     (VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT | VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) != 0) {
+                    renderbufferCaps |= GetAttachmentCaps(logicalFormat);
+                    renderbufferCaps |= FormatCapability::MultisampleRenderbuffer;
+                }
+                cache.FullCaps[renderbufferTargetIndex][formatIndex] |= renderbufferCaps;
+
+                if (fallbackFormat != VK_FORMAT_UNDEFINED && fallbackFormat != nativeFormat) {
+                    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)) !=
+                        0) {
+                        fallbackRenderbufferCaps |= GetAttachmentCaps(logicalFormat);
+                        fallbackRenderbufferCaps |= FormatCapability::MultisampleRenderbuffer;
+                    }
+                    cache.CaveatCaps[renderbufferTargetIndex][formatIndex] |= fallbackRenderbufferCaps;
+                    if (fallbackLogicalFormat &&
+                        HasNewCaveatFormatCaps(renderbufferCaps, fallbackRenderbufferCaps)) {
+                        LogVulkanFormatCaveat(logicalFormat, renderbufferTargetIndex, *fallbackLogicalFormat);
+                    }
+                }
+
+                const FormatCapabilityFlags rbCaps = cache.FullCaps[renderbufferTargetIndex][formatIndex] |
+                                                     cache.CaveatCaps[renderbufferTargetIndex][formatIndex];
+                if (HasFormatCapability(rbCaps, FormatCapability::MultisampleRenderbuffer)) {
+                    cache.SampleCounts[renderbufferTargetIndex][formatIndex] =
+                        BuildSampleCounts(capabilities.MaxFramebufferSamples);
+                }
+            }
         }
     } // namespace
 
+    void PopulateFormatCapabilities(VkPhysicalDevice physicalDevice,
+                                    PFN_vkGetPhysicalDeviceFormatProperties getFormatProperties,
+                                    const MG_External::VulkanCapabilities& capabilities,
+                                    FormatCapabilityCache& cache) {
+        PopulateFormatCapabilitiesImpl(physicalDevice, getFormatProperties, capabilities, cache);
+    }
+
     BackendObject_DirectVulkan::~BackendObject_DirectVulkan() = default;
 
+    BackendObject_DirectVulkan::BackendObject_DirectVulkan(): m_rendererInfo{GetRendererIdentity()} {}
+
     Bool BackendObject_DirectVulkan::InitWindowSurface() {
         if (!m_windowHandle.Handle) {
             MGLOG_E("Cannot initialize DirectVulkan window surface: native window handle is null");
@@ -27,12 +345,28 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         auto nativeWindow = reinterpret_cast<NativeWindowType>(m_windowHandle.Handle);
 
+        // Any renderer instance this assignment replaces is destroyed here;
+        // fence/timer-query handles stamped with the old generation go stale.
+        BumpRendererGeneration();
         pVulkanRenderer = MakeUnique<MG_Backend::DirectVulkan::VulkanRenderer>(nativeWindow);
         MOBILEGL_ASSERT(pVulkanRenderer != nullptr, "InitWindowSurface: VulkanRenderer creation failed");
         pVulkanRenderer->Initialize();
         return true;
     }
 
+    Bool BackendObject_DirectVulkan::InitPbufferSurface(EGLint width, EGLint height) {
+        VulkanRendererConfig config;
+        config.SurfaceWidth = static_cast<Uint32>(std::max<EGLint>(width, 1));
+        config.SurfaceHeight = static_cast<Uint32>(std::max<EGLint>(height, 1));
+        // Any renderer instance this assignment replaces is destroyed here;
+        // fence/timer-query handles stamped with the old generation go stale.
+        BumpRendererGeneration();
+        pVulkanRenderer = MakeUnique<MG_Backend::DirectVulkan::VulkanRenderer>(NativeWindowType{}, config);
+        MOBILEGL_ASSERT(pVulkanRenderer != nullptr, "InitPbufferSurface: VulkanRenderer creation failed");
+        pVulkanRenderer->Initialize();
+        return true;
+    }
+
     void BackendObject_DirectVulkan::Initialize() {
         m_initialized = true;
     }
@@ -47,8 +381,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return false;
         }
 
-        MG_Util::BackendLoader::FillInVulkanCapabilities(m_vulkanCaps, pVulkanRenderer->GetPhysicalDevice().properties);
+        const auto& physicalDevice = pVulkanRenderer->GetPhysicalDevice();
+        if (!MG_Util::BackendLoader::QueryVulkanCapabilities(m_vulkanCaps, pVulkanRenderer->GetInstance(),
+                                                             physicalDevice.handle)) {
+            MGLOG_W("DirectVulkan: failed to query extended Vulkan capabilities, using basic properties");
+            MG_Util::BackendLoader::FillInVulkanCapabilities(m_vulkanCaps, physicalDevice.properties);
+        }
         UpdateDynamicBackendParameters();
+        UpdateAdvertisedExtensions();
+        PopulateFormatCapabilities(physicalDevice.handle, vkGetPhysicalDeviceFormatProperties, m_vulkanCaps,
+                                   MutableFormatCapabilities());
+        PrintFormatCapabilities(GetFormatCapabilities());
         return true;
     }
 
@@ -60,40 +403,49 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         return BackendObject::InitializeEGLDisplay(dpy, major, minor);
     }
 
-    Bool BackendObject_DirectVulkan::CreateEGLWindowSurface(const WindowHandle& handle) {
+    Bool BackendObject_DirectVulkan::CreateEGLWindowSurface(EGLSurface surface, const WindowHandle& handle) {
         const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
         if (!m_initialized) {
             MGLOG_E("DirectVulkan backend not initialized");
             return false;
         }
-        if (handle.Backend != WindowBackend::Android || !handle.Handle) {
-            MGLOG_E("DirectVulkan backend only supports Android 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;
         }
 
-        const Bool sameHandle =
-            m_eglWindowSurfaceInitialized && m_windowHandle.Backend == handle.Backend && m_windowHandle.Handle == handle.Handle;
-        if (sameHandle) {
-            return true;
-        }
+        return RegisterEGLWindowSurface(surface, handle);
+    }
 
-        if (m_eglWindowSurfaceInitialized || pVulkanRenderer) {
-            pVulkanRenderer.reset();
-            ResetEGLRuntimeState();
+    Bool BackendObject_DirectVulkan::ResizeEGLWindowSurface(EGLSurface surface, Uint32 width, Uint32 height) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        if (!m_initialized) {
+            MGLOG_E("DirectVulkan backend not initialized");
+            return false;
         }
+        if (!BackendObject::ResizeEGLWindowSurface(surface, width, height)) {
+            return false;
+        }
+        if (pVulkanRenderer && m_eglSurface == surface) {
+            pVulkanRenderer->RequestSwapchainResize(width, height);
+        }
+        return true;
+    }
 
-        return BackendObject::CreateEGLWindowSurface(handle);
+    Bool BackendObject_DirectVulkan::CreateEGLPbufferSurface(EGLSurface surface, EGLint width, EGLint height) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        if (!m_initialized) {
+            MGLOG_E("DirectVulkan backend not initialized");
+            return false;
+        }
+        return RegisterEGLPbufferSurface(surface, width, height);
     }
 
     Bool BackendObject_DirectVulkan::MakeEGLCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx) {
         const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
-        if (IsReleaseCurrentRequest(dpy, draw, read, ctx)) {
-            return BackendObject::MakeEGLCurrent(dpy, draw, read, ctx);
-        }
-        if (!pVulkanRenderer) {
-            MGLOG_E("DirectVulkan renderer is not initialized");
-            return false;
-        }
         return BackendObject::MakeEGLCurrent(dpy, draw, read, ctx);
     }
 
@@ -106,33 +458,95 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         return BackendObject::SwapEGLBuffers(dpy, draw);
     }
 
+    void BackendObject_DirectVulkan::ReleaseEGLSurface(EGLSurface surface) {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        BackendObject::ReleaseEGLSurface(surface);
+    }
+
+    void BackendObject_DirectVulkan::ReleaseEGLResources() {
+        const std::lock_guard<std::recursive_mutex> lock(m_eglStateMutex);
+        // Outstanding fence/timer-query handles now refer to a dead renderer;
+        // treat them as signaled/available with zero results from here on.
+        BumpRendererGeneration();
+        pVulkanRenderer.reset();
+        BackendObject::ReleaseEGLResources();
+    }
+
+    void BackendObject_DirectVulkan::OnEGLSurfaceReleased(EGLSurface surface) {
+        (void)surface;
+        // Outstanding fence/timer-query handles now refer to a dead renderer;
+        // treat them as signaled/available with zero results from here on.
+        BumpRendererGeneration();
+        pVulkanRenderer.reset();
+    }
+
     const RendererInfo& BackendObject_DirectVulkan::GetRendererInfo() const {
-        static RendererInfo RendererInfo = {
-            .RendererName = "Magma",          // Renderer Name
-            .BackendName = "Direct (Vulkan)", // Backend Name
-            .ExtraVendor = Nullopt,           // Extra vendor
-            .RendererGLInfo =
-                {
-                    .TargetGLVersion = {3, 3, 0},                      // Target OpenGL Version
-                    .TargetGLSLVersion = {4, 6, 0},                    // Target Shading Language Version
-                    .Extensions = {V_OpenGL30, V_OpenGL31, V_OpenGL32, // OpenGL Extensions
-                                   V_OpenGL33},
-                    .IsCompatibilityProfile = false // Is Compatibility Profile
-                },
-            .StaticBackendCapability = {.AllowVSOnlyPrograms = false} // Backend Capability
-        };
-        return RendererInfo;
+        return m_rendererInfo;
     }
 
     String BackendObject_DirectVulkan::GetBackendAPIVersionString() const {
         if (!m_initialized) {
             return "<uninitialized DirectVulkan backend>";
         }
+        return FormatBackendAPIVersionString(m_vulkanCaps.DeviceName, m_vulkanCaps.VulkanAPIVersion.toString(),
+                                             m_vulkanCaps.DriverVersionString);
+    }
+
+    const RendererInfo& GetRendererIdentity() {
+        static const RendererInfo rendererInfo = {
+            .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
+                },
+            .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) {
+            extensions.push_back(E_GL_KHR_shader_subgroup);
+        }
+        // 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.
+        if (timerQueriesSupported && !MG_Config::Features.DisableTimerQuery) {
+            extensions.push_back(E_GL_ARB_timer_query);
+        }
+        // Only advertised when the samplerAnisotropy device feature was granted: without it the
+        // sampler state is accepted but never applied, and an app trusting the string (LWJGL builds
+        // GLCapabilities from it) would think it enabled anisotropic filtering.
+        if (anisotropicFilteringSupported) {
+            extensions.push_back(E_GL_EXT_texture_filter_anisotropic);
+            extensions.push_back(E_GL_ARB_texture_filter_anisotropic);
+        }
+        return extensions;
+    }
+
+    String FormatBackendAPIVersionString(const String& deviceName, const String& vulkanApiVersionString,
+                                         const String& driverVersionString) {
         // Format:
         // <GPU Name>, Vulkan <Vulkan Version>, Driver <Driver Version>
-        String str = m_vulkanCaps.DeviceName + ", Vulkan " + m_vulkanCaps.VulkanAPIVersion.toString() + ", Driver " +
-                     m_vulkanCaps.DriverVersionString;
-        return str;
+        return deviceName + ", Vulkan " + vulkanApiVersionString + ", Driver " + driverVersionString;
     }
 
     BackendType BackendObject_DirectVulkan::GetBackendType() const {
@@ -147,10 +561,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             funcsTable.GL.DrawArrays = DrawArrays;
             funcsTable.GL.DrawElements = DrawElements;
             funcsTable.GL.DrawElementsBaseVertex = DrawElementsBaseVertex;
+            funcsTable.GL.MultiDrawArrays = MultiDrawArrays;
             funcsTable.GL.MultiDrawElements = MultiDrawElements;
             funcsTable.GL.MultiDrawElementsBaseVertex = MultiDrawElementsBaseVertex;
             funcsTable.GL.MultiDrawElementsIndirect = MultiDrawElementsIndirect;
             funcsTable.GL.MultiDrawArraysIndirect = MultiDrawArraysIndirect;
+            funcsTable.GL.MultiDrawElementsIndirectCount = MultiDrawElementsIndirectCount;
+            funcsTable.GL.MultiDrawArraysIndirectCount = MultiDrawArraysIndirectCount;
             funcsTable.GL.DrawRangeElementsBaseVertex = DrawRangeElementsBaseVertex;
             funcsTable.GL.DrawRangeElements = DrawRangeElements;
             funcsTable.GL.DrawElementsInstancedBaseVertexBaseInstance = DrawElementsInstancedBaseVertexBaseInstance;
@@ -166,12 +583,51 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             funcsTable.GL.ClearBufferfv = ClearBufferfv;
             funcsTable.GL.ClearBufferuiv = ClearBufferuiv;
             funcsTable.GL.ClearBufferiv = ClearBufferiv;
+            funcsTable.GL.ClearNamedFramebufferfv = ClearNamedFramebufferfv;
+            funcsTable.GL.ClearNamedFramebufferfi = ClearNamedFramebufferfi;
             funcsTable.GL.BlitFramebuffer = BlitFramebuffer;
+            funcsTable.GL.BlitNamedFramebuffer = BlitNamedFramebuffer;
             funcsTable.GL.CopyTexImage2D = CopyTexImage2D;
             funcsTable.GL.CopyTexSubImage2D = CopyTexSubImage2D;
+            funcsTable.GL.CopyImageSubData = CopyImageSubData;
             funcsTable.GL.GenerateMipmap = GenerateMipmap;
             funcsTable.GL.ReadPixels = ReadPixels;
             funcsTable.GL.GetTexImage = GetTexImage;
+            funcsTable.GL.GetTextureImage = GetTextureImage;
+            funcsTable.GL.DispatchCompute = DispatchCompute;
+            funcsTable.GL.DispatchComputeIndirect = DispatchComputeIndirect;
+            funcsTable.GL.MemoryBarrier = MemoryBarrier;
+            funcsTable.GL.MemoryBarrierByRegion = MemoryBarrierByRegion;
+            funcsTable.GL.BindImageTexture = BindImageTexture;
+            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;
+            funcsTable.GL.WaitSync = WaitSync;
+            funcsTable.GL.DeleteSync = DeleteSync;
+            funcsTable.GL.GetSyncStatus = GetSyncStatus;
+            // Optional timer-query group: left null (the frontend then falls
+            // back) when disabled via MOBILEGL_DISABLE_TIMERQUERY. The hooks
+            // themselves additionally degrade to null handles when the device
+            // lacks timestamp support.
+            if (!MG_Config::Features.DisableTimerQuery) {
+                funcsTable.GL.IsTimerQuerySupported = IsTimerQuerySupported;
+                funcsTable.GL.BeginTimeElapsedQuery = BeginTimeElapsedQuery;
+                funcsTable.GL.EndTimeElapsedQuery = EndTimeElapsedQuery;
+                funcsTable.GL.QueryCounterTimestamp = QueryCounterTimestamp;
+                funcsTable.GL.IsQueryResultAvailable = IsQueryResultAvailable;
+                funcsTable.GL.GetQueryResult64 = GetQueryResult64;
+                funcsTable.GL.DeleteBackendQuery = DeleteBackendQuery;
+                funcsTable.GL.GetGpuTimestampNs = GetGpuTimestampNs;
+            }
             funcsTableInitialized = true;
         }
         return funcsTable;
@@ -181,7 +637,204 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         return m_dynamicParameters;
     }
 
+    void BackendObject_DirectVulkan::ApplyVulkanCapabilitiesForTesting(
+        const MG_External::VulkanCapabilities& capabilities) {
+        m_vulkanCaps = capabilities;
+        UpdateDynamicBackendParameters();
+        UpdateAdvertisedExtensions();
+        MutableFormatCapabilities().Clear();
+    }
+
+    void BackendObject_DirectVulkan::UpdateAdvertisedExtensions() {
+        // GL_ARB_timer_query gates MC's F3 GPU% (LWJGL checks the extension
+        // string). InitCapabilities runs after InitWindowSurface has created
+        // and initialized the renderer, so the advertisement can be gated on
+        // real device timestamp support. ApplyVulkanCapabilitiesForTesting may
+        // run without a renderer; no timer query is advertised then. Rebuilding
+        // the whole list keeps re-runs idempotent.
+        m_rendererInfo.RendererGLInfo.Extensions = BuildAdvertisedExtensions(
+            m_vulkanCaps.SupportsShaderSubgroup, pVulkanRenderer && pVulkanRenderer->IsTimerQuerySupported(),
+            pVulkanRenderer && pVulkanRenderer->IsSamplerAnisotropySupported());
+    }
+
     void BackendObject_DirectVulkan::UpdateDynamicBackendParameters() {
+        const auto mapShaderStages = [](Uint32 vkStages) {
+            Uint32 glStages = 0;
+            if ((vkStages & VK_SHADER_STAGE_VERTEX_BIT) != 0) glStages |= GL_VERTEX_SHADER_BIT;
+            if ((vkStages & VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT) != 0) glStages |= GL_TESS_CONTROL_SHADER_BIT;
+            if ((vkStages & VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT) != 0) {
+                glStages |= GL_TESS_EVALUATION_SHADER_BIT;
+            }
+            if ((vkStages & VK_SHADER_STAGE_GEOMETRY_BIT) != 0) glStages |= GL_GEOMETRY_SHADER_BIT;
+            if ((vkStages & VK_SHADER_STAGE_FRAGMENT_BIT) != 0) glStages |= GL_FRAGMENT_SHADER_BIT;
+            if ((vkStages & VK_SHADER_STAGE_COMPUTE_BIT) != 0) glStages |= GL_COMPUTE_SHADER_BIT;
+            return glStages;
+        };
+
+        const auto mapSubgroupFeatures = [](Uint32 vkFeatures) {
+            Uint32 glFeatures = 0;
+            if ((vkFeatures & VK_SUBGROUP_FEATURE_BASIC_BIT) != 0) {
+                glFeatures |= GL_SUBGROUP_FEATURE_BASIC_BIT_KHR;
+            }
+            if ((vkFeatures & VK_SUBGROUP_FEATURE_VOTE_BIT) != 0) {
+                glFeatures |= GL_SUBGROUP_FEATURE_VOTE_BIT_KHR;
+            }
+            if ((vkFeatures & VK_SUBGROUP_FEATURE_ARITHMETIC_BIT) != 0) {
+                glFeatures |= GL_SUBGROUP_FEATURE_ARITHMETIC_BIT_KHR;
+            }
+            if ((vkFeatures & VK_SUBGROUP_FEATURE_BALLOT_BIT) != 0) {
+                glFeatures |= GL_SUBGROUP_FEATURE_BALLOT_BIT_KHR;
+            }
+            if ((vkFeatures & VK_SUBGROUP_FEATURE_SHUFFLE_BIT) != 0) {
+                glFeatures |= GL_SUBGROUP_FEATURE_SHUFFLE_BIT_KHR;
+            }
+            if ((vkFeatures & VK_SUBGROUP_FEATURE_SHUFFLE_RELATIVE_BIT) != 0) {
+                glFeatures |= GL_SUBGROUP_FEATURE_SHUFFLE_RELATIVE_BIT_KHR;
+            }
+            if ((vkFeatures & VK_SUBGROUP_FEATURE_CLUSTERED_BIT) != 0) {
+                glFeatures |= GL_SUBGROUP_FEATURE_CLUSTERED_BIT_KHR;
+            }
+            if ((vkFeatures & VK_SUBGROUP_FEATURE_QUAD_BIT) != 0) {
+                glFeatures |= GL_SUBGROUP_FEATURE_QUAD_BIT_KHR;
+            }
+            return glFeatures;
+        };
+
+        static constexpr SizeT kMaxAdvertisedShaderStorageBlockSize = 512ull * 1024ull * 1024ull;
         m_dynamicParameters.UniformBufferOffsetAlignment = m_vulkanCaps.UniformBufferOffsetAlignment;
+        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;
+        m_dynamicParameters.SmoothLineWidthRangeMin = m_vulkanCaps.SmoothLineWidthRangeMin;
+        m_dynamicParameters.SmoothLineWidthRangeMax = m_vulkanCaps.SmoothLineWidthRangeMax;
+        m_dynamicParameters.SmoothLineWidthGranularity = m_vulkanCaps.SmoothLineWidthGranularity;
+        m_dynamicParameters.PointSizeRangeMin = m_vulkanCaps.PointSizeRangeMin;
+        m_dynamicParameters.PointSizeRangeMax = m_vulkanCaps.PointSizeRangeMax;
+        m_dynamicParameters.PointSizeGranularity = m_vulkanCaps.PointSizeGranularity;
+        m_dynamicParameters.Max3DTextureSize = m_vulkanCaps.Max3DTextureSize;
+        m_dynamicParameters.MaxArrayTextureLayers = m_vulkanCaps.MaxArrayTextureLayers;
+        m_dynamicParameters.MaxCubeMapTextureSize = m_vulkanCaps.MaxCubeMapTextureSize;
+        m_dynamicParameters.MaxFramebufferWidth = m_vulkanCaps.MaxFramebufferWidth;
+        m_dynamicParameters.MaxFramebufferHeight = m_vulkanCaps.MaxFramebufferHeight;
+        m_dynamicParameters.MaxFramebufferLayers = m_vulkanCaps.MaxFramebufferLayers;
+        m_dynamicParameters.MaxRenderbufferSize = m_vulkanCaps.MaxRenderbufferSize;
+        m_dynamicParameters.MaxTextureSize = m_vulkanCaps.MaxTextureSize;
+        m_dynamicParameters.MaxColorTextureSamples = m_vulkanCaps.MaxColorTextureSamples;
+        m_dynamicParameters.MaxDepthTextureSamples = m_vulkanCaps.MaxDepthTextureSamples;
+        m_dynamicParameters.MaxFramebufferSamples = m_vulkanCaps.MaxFramebufferSamples;
+        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);
+        // 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
+        // an array bound: Minecraft's Blaze3D GlStateManager.TEXTURES[] holds 128 entries and Iris
+        // iterates [0, GL_MAX_TEXTURE_IMAGE_UNITS) over it (CompositeRenderer.renderAll), so any value
+        // > 128 throws ArrayIndexOutOfBoundsException. Match desktop drivers (32) for the per-stage
+        // 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.MaxVertexTextureImageUnits =
+            std::min(m_vulkanCaps.MaxVertexTextureImageUnits, maxPerStageTextureUnits);
+        m_dynamicParameters.MaxComputeTextureImageUnits =
+            std::min(m_vulkanCaps.MaxComputeTextureImageUnits, maxPerStageTextureUnits);
+        m_dynamicParameters.MaxCombinedTextureImageUnits =
+            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.MaxComputeWorkGroupInvocations = m_vulkanCaps.MaxComputeWorkGroupInvocations;
+        m_dynamicParameters.MaxShaderStorageBufferBindings = m_vulkanCaps.MaxShaderStorageBufferBindings;
+        m_dynamicParameters.MaxTextureBufferSize = m_vulkanCaps.MaxTextureBufferSize;
+        m_dynamicParameters.MaxUniformBufferBindings = m_vulkanCaps.MaxUniformBufferBindings;
+        m_dynamicParameters.MaxUniformBlockSize = m_vulkanCaps.MaxUniformBlockSize;
+        m_dynamicParameters.MaxImageUnits =
+            std::max(std::min(m_vulkanCaps.MaxImageUnits, maxSupportedTextureUnits), 0);
+        m_dynamicParameters.MaxCombinedImageUniforms = std::max(m_vulkanCaps.MaxCombinedImageUniforms, 0);
+        const Int maxPerStageImageUniforms =
+            std::min(m_dynamicParameters.MaxImageUnits, m_dynamicParameters.MaxCombinedImageUniforms);
+        // Vulkan uses one descriptor limit for every stage, but non-compute stores/atomics are
+        // optional device features. VulkanRenderer enables each feature whenever the physical
+        // device reports it, so these are the exact limits the logical device can compile and run.
+        m_dynamicParameters.MaxVertexImageUniforms =
+            m_vulkanCaps.SupportsVertexPipelineStoresAndAtomics ? maxPerStageImageUniforms : 0;
+        m_dynamicParameters.MaxGeometryImageUniforms =
+            m_vulkanCaps.SupportsVertexPipelineStoresAndAtomics && m_vulkanCaps.SupportsGeometryShader
+                ? maxPerStageImageUniforms
+                : 0;
+        m_dynamicParameters.MaxFragmentImageUniforms =
+            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);
+        m_dynamicParameters.MaxDrawBuffers = std::min(m_vulkanCaps.MaxDrawBuffers, maxSupportedDrawBuffers);
+        m_dynamicParameters.MaxColorAttachments = std::min(m_vulkanCaps.MaxColorAttachments, maxSupportedDrawBuffers);
+        m_dynamicParameters.MaxClipDistances = m_vulkanCaps.MaxClipDistances;
+        m_dynamicParameters.MaxViewports = m_vulkanCaps.MaxViewports;
+        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.SupportsWideLines = m_vulkanCaps.SupportsWideLines;
+        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.SubgroupQuadOperationsInAllStages = m_vulkanCaps.SubgroupQuadOperationsInAllStages;
+        } else {
+            m_dynamicParameters.SubgroupSize = 0;
+            m_dynamicParameters.SubgroupSupportedStages = 0;
+            m_dynamicParameters.SubgroupSupportedFeatures = 0;
+            m_dynamicParameters.SubgroupQuadOperationsInAllStages = false;
+        }
+        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);
+        }
+        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

@@ -12,29 +12,73 @@
 #include <MG_Util/BackendLoaders/Vulkan/Loader.h>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
+    // Populates the same format-capability cache used by backend startup. Passing the
+    // instance-resolved function keeps standalone callers independent of global loader
+    // initialization; the physical device must remain valid for the duration of the call.
+    void PopulateFormatCapabilities(VkPhysicalDevice physicalDevice,
+                                    PFN_vkGetPhysicalDeviceFormatProperties getFormatProperties,
+                                    const MG_External::VulkanCapabilities& capabilities,
+                                    FormatCapabilityCache& cache);
+
     class BackendObject_DirectVulkan : public BackendObject {
     public:
+        BackendObject_DirectVulkan();
         ~BackendObject_DirectVulkan() override;
 
         void Initialize() override;
         Bool InitWindowSurface() override;
         Bool InitCapabilities() override;
         Bool InitializeEGLDisplay(EGLDisplay dpy, EGLint* major, EGLint* minor) override;
-        Bool CreateEGLWindowSurface(const WindowHandle& handle) override;
+        Bool CreateEGLWindowSurface(EGLSurface surface, const WindowHandle& handle) override;
+        Bool ResizeEGLWindowSurface(EGLSurface surface, Uint32 width, Uint32 height) override;
+        Bool CreateEGLPbufferSurface(EGLSurface surface, EGLint width, EGLint height) override;
         Bool MakeEGLCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx) override;
         Bool SwapEGLBuffers(EGLDisplay dpy, EGLSurface draw) override;
+        void ReleaseEGLSurface(EGLSurface surface) override;
+        void ReleaseEGLResources() override;
 
         const RendererInfo& GetRendererInfo() const override;
         String GetBackendAPIVersionString() const override;
         const GlobalBackendFunctionsTable& GetBackendFunctions() const override;
         const DynamicBackendParameters& GetDynamicParameters() const override;
         BackendType GetBackendType() const override;
+        void ApplyVulkanCapabilitiesForTesting(const MG_External::VulkanCapabilities& capabilities);
 
     private:
+        Bool InitPbufferSurface(EGLint width, EGLint height) override;
+        void OnEGLSurfaceReleased(EGLSurface surface) override;
+        void UpdateAdvertisedExtensions();
         void UpdateDynamicBackendParameters();
 
         Bool m_initialized = false;
         DynamicBackendParameters m_dynamicParameters;
         MG_External::VulkanCapabilities m_vulkanCaps;
+        RendererInfo m_rendererInfo;
     };
+
+    // Single-source-of-truth helpers shared with the driver POST
+    // (MG_Util/SelfTest/DriverPost.cpp), so the identity strings and extension list
+    // MobileGL reports to applications on this backend cannot drift from what the
+    // 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
+    // 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
+    // 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);
+
+    // Format: <GPU Name>, Vulkan <Vulkan Version>, Driver <Driver Version> — the exact
+    // string an initialized backend returns from GetBackendAPIVersionString (and that
+    // ends up inside the application-visible GL_RENDERER string).
+    String FormatBackendAPIVersionString(const String& deviceName, const String& vulkanApiVersionString,
+                                         const String& driverVersionString);
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/DirectVulkan.h

@@ -8,25 +8,44 @@
 
 #pragma once
 #include <Includes.h>
+#include <MG_Backend/BackendObject.h>
 #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
+    // pVulkanRenderer is reset or recreated; fence and timer-query handles
+    // stamped with an older generation are stale and resolve as signaled /
+    // available with zero results instead of dereferencing the destroyed
+    // renderer's frame serials and query-pool slots.
+    Uint64 GetRendererGeneration();
+    void BumpRendererGeneration();
 
     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 ClearNamedFramebufferfi(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer, GLenum buffer,
+                                 GLint drawbuffer, GLfloat depth, GLint stencil);
     void Clear(GLbitfield mask);
     void DrawElements(GLenum mode, GLsizei count, GLenum type, const void* indices);
     void DrawArrays(GLenum mode, GLint first, GLsizei count);
     void DrawElementsBaseVertex(GLenum mode, GLsizei count, GLenum type, const GLvoid* indices, GLint basevertex);
+    void MultiDrawArrays(GLenum mode, const GLint* first, const GLsizei* count, GLsizei drawcount);
     void MultiDrawElements(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
                            GLsizei drawcount);
     void MultiDrawElementsBaseVertex(GLenum mode, const GLsizei* count, GLenum type, const GLvoid* const* indices,
                                      GLsizei drawcount, const GLint* basevertex);
     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,
+                                        GLsizei maxdrawcount, 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);
@@ -44,12 +63,72 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     void DrawArraysIndirect(GLenum mode, const void* indirect);
     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,
+                              const SharedPtr<MG_State::GLState::FramebufferObject>& drawFramebuffer,
+                              GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                              GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                              GLbitfield mask, GLenum filter);
     void CopyTexImage2D(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width,
                         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,
+                          GLenum srcTarget, GLint srcLevel, GLint srcX, GLint srcY, GLint srcZ,
+                          const SharedPtr<MG_State::GLState::ITextureObject>& dstTexture,
+                          GLenum dstTarget, GLint dstLevel, GLint dstX, GLint dstY, GLint dstZ,
+                          GLsizei srcWidth, GLsizei srcHeight, GLsizei srcDepth);
     void GenerateMipmap(GLenum target);
+    void DispatchCompute(GLuint numGroupsX, GLuint numGroupsY, GLuint numGroupsZ);
+    void DispatchComputeIndirect(GLintptr indirect);
+    void MemoryBarrier(GLbitfield barriers);
+    void MemoryBarrierByRegion(GLbitfield barriers);
+    void BindImageTexture(GLuint unit, GLuint texture, GLint level, GLboolean layered, GLint layer, GLenum access,
+                          GLenum format);
+    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 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,
+                         GLint level, GLenum format, GLenum type, GLsizei bufSize, GLvoid* pixels);
+    // GL fence sync objects, mapped onto the renderer's frame-serial busy
+    // tracking: a fence captures the frame serial current at creation and is
+    // signaled once every command recorded under that serial has completed on
+    // the GPU.
+    BackendSyncHandle FenceSync();
+    GLenum ClientWaitSync(BackendSyncHandle sync, GLbitfield flags, GLuint64 timeout);
+    void WaitSync(BackendSyncHandle sync, GLbitfield flags, GLuint64 timeout);
+    void DeleteSync(BackendSyncHandle sync);
+    Bool GetSyncStatus(BackendSyncHandle sync);
+    // GPU timer queries (GL_TIME_ELAPSED spans and GL_TIMESTAMP one-shots),
+    // backed by per-frame VkQueryPool timestamp slots. All hooks degrade
+    // gracefully: null handles when the renderer is absent, the device lacks
+    // timestamp support, or the frame's pool is exhausted.
+    // Dynamic support check (GLFunctionsTable::IsTimerQuerySupported): true
+    // only while a live renderer exists whose device can actually time.
+    Bool IsTimerQuerySupported();
+    BackendQueryHandle BeginTimeElapsedQuery();
+    void EndTimeElapsedQuery(BackendQueryHandle query);
+    BackendQueryHandle QueryCounterTimestamp();
+    Bool IsQueryResultAvailable(BackendQueryHandle query);
+    // Returns true when a final value was produced (outNanoseconds set; the
+    // frontend may cache it and release the handle), false when the result
+    // cannot be obtained yet (e.g. a wait refused because the records' frame
+    // serial is the current unsubmitted frame) - the handle then stays
+    // readable later.
+    Bool GetQueryResult64(BackendQueryHandle query, Bool wait, Uint64* outNanoseconds);
+    void DeleteBackendQuery(BackendQueryHandle query);
+    // Always 0: Vulkan cannot synchronously sample the GPU clock (timestamps
+    // only exist as vkCmdWriteTimestamp results); the frontend falls back.
+    Int64 GetGpuTimestampNs();
     void Present();
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/DirectVulkanResourceState.h

@@ -0,0 +1,20 @@
+// MobileGL - MobileGL/MG_Backend/DirectVulkan/DirectVulkanResourceState.h
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#pragma once
+
+#include <Includes.h>
+
+namespace MobileGL::MG_State::GLState {
+    class ProgramObject;
+}
+
+namespace MobileGL::MG_Backend::DirectVulkan {
+    GLuint GetShaderStorageBlockIndex(const MG_State::GLState::ProgramObject& program, const String& name);
+    GLuint GetShaderStorageBlockBinding(const MG_State::GLState::ProgramObject& program, GLuint blockIndex);
+}

MobileGL/MG_Backend/DirectVulkan/Renderer/FrameContext.cpp

@@ -13,6 +13,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Destroy(device, commandPool);
         m_frames.assign(frameCount, {});
         currentFrameIndex = 0;
+        m_device = device;
+        m_commandPool = commandPool;
 
         Vector<VkCommandBuffer> commandBuffers(frameCount, VK_NULL_HANDLE);
         VkCommandBufferAllocateInfo allocInfo{};
@@ -55,10 +57,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
         DestroySwapchainSemaphores(device);
         if (device != VK_NULL_HANDLE && commandPool != VK_NULL_HANDLE && !m_frames.empty()) {
+            for (auto& frame : m_frames) {
+                FreeRetiredCommandBuffers(frame);
+            }
             vkFreeCommandBuffers(device, commandPool, frameCount, commandBuffers.data());
         }
         m_frames.clear();
         currentFrameIndex = 0;
+        m_device = VK_NULL_HANDLE;
+        m_commandPool = VK_NULL_HANDLE;
     }
 
     FrameContext::FrameData& FrameContext::GetCurrent() {
@@ -97,6 +104,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VK_VERIFY(vkBeginCommandBuffer(frame.commandBuffer, &beginInfo), "BeginCommandRecording, vkBeginCommandBuffer");
 
         frame.isCommandRecording = true;
+        if (m_recordingObserver != nullptr) {
+            m_recordingObserver->OnFrameCommandRecordingBegan(frame.commandBuffer);
+        }
         return frame.commandBuffer;
     }
 
@@ -178,9 +188,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         packet.signalSemaphore = m_swapchainImageRenderFinishedSemaphores[swapchainImageIndex];
         packet.commandBuffer = frame.commandBuffer;
 
-        packet.submitInfo.waitSemaphoreCount = 1;
-        packet.submitInfo.pWaitSemaphores = &packet.waitSemaphore;
-        packet.submitInfo.pWaitDstStageMask = &packet.waitDstStageMask;
+        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.signalSemaphoreCount = 1;
@@ -188,18 +198,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         return packet;
     }
 
-    FrameContext::PresentInfoPacket FrameContext::GetPresentInfo(VkSwapchainKHR swapchain, const Uint32& imageIndex) const {
+    FrameContext::PresentInfoPacket FrameContext::GetPresentInfo(VkSwapchainKHR swapchain, Uint32 imageIndex) const {
         AssertValidSwapchainImageIndex(imageIndex);
         PresentInfoPacket packet{};
         packet.waitSemaphore = m_swapchainImageRenderFinishedSemaphores[imageIndex];
         packet.swapchain = swapchain;
-        packet.imageIndex = &imageIndex;
+        packet.imageIndex = imageIndex;
 
         packet.presentInfo.waitSemaphoreCount = 1;
         packet.presentInfo.pWaitSemaphores = &packet.waitSemaphore;
         packet.presentInfo.swapchainCount = 1;
         packet.presentInfo.pSwapchains = &packet.swapchain;
-        packet.presentInfo.pImageIndices = packet.imageIndex;
+        packet.presentInfo.pImageIndices = &packet.imageIndex;
         packet.presentInfo.pResults = nullptr;
         return packet;
     }
@@ -211,14 +221,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         if (result != VK_SUCCESS) {
             return result;
         }
+        // The slot's fence has been waited: every command buffer this slot
+        // submitted (including mid-frame flushes) has finished executing.
+        FreeRetiredCommandBuffers(frame);
 
-        result = vkResetFences(device, 1, &frame.imageInFlightFence);
+        result = vkAcquireNextImageKHR(device, swapchain, timeout, frame.imageAvailableSemaphore, acquireFence,
+                                       &outImageIndex);
         if (result != VK_SUCCESS) {
             return result;
         }
 
-        return vkAcquireNextImageKHR(device, swapchain, timeout, frame.imageAvailableSemaphore, acquireFence,
-                                     &outImageIndex);
+        frame.imageAvailableSemaphoreConsumed = false;
+        return vkResetFences(device, 1, &frame.imageInFlightFence);
     }
 
     Uint32 FrameContext::GetCurrentFrameIndex() const {
@@ -229,6 +243,43 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         return static_cast<Uint32>(m_frames.size());
     }
 
+    void FrameContext::SetRecordingObserver(IRecordingObserver* observer) {
+        m_recordingObserver = observer;
+    }
+
+    VkResult FrameContext::RetireCurrentCommandBuffer() {
+        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");
+
+        VkCommandBufferAllocateInfo allocInfo{};
+        allocInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
+        allocInfo.commandPool = m_commandPool;
+        allocInfo.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
+        allocInfo.commandBufferCount = 1;
+        VkCommandBuffer replacement = VK_NULL_HANDLE;
+        const VkResult result = vkAllocateCommandBuffers(m_device, &allocInfo, &replacement);
+        if (result != VK_SUCCESS) {
+            return result;
+        }
+        frame.retiredCommandBuffers.push_back(frame.commandBuffer);
+        frame.commandBuffer = replacement;
+        return VK_SUCCESS;
+    }
+
+    void FrameContext::FreeRetiredCommandBuffers(FrameData& frame) {
+        if (frame.retiredCommandBuffers.empty()) {
+            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());
+        }
+        frame.retiredCommandBuffers.clear();
+    }
+
     void FrameContext::AssertValidFrameIndex(Uint32 frameIndex) const {
         MOBILEGL_ASSERT(frameIndex < m_frames.size(), "FrameContext index out of range");
     }
@@ -260,6 +311,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         frame.hasCommandBufferRecorded = false;
         frame.isCommandRecording = false;
+        frame.imageAvailableSemaphoreConsumed = false;
         return VK_SUCCESS;
     }
 
@@ -277,5 +329,6 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         frame.imageAvailableSemaphore = VK_NULL_HANDLE;
         frame.isCommandRecording = false;
         frame.hasCommandBufferRecorded = false;
+        frame.imageAvailableSemaphoreConsumed = false;
     }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/FrameContext.h

@@ -14,6 +14,17 @@
 namespace MobileGL::MG_Backend::DirectVulkan {
     class FrameContext {
     public:
+        // Notified immediately after a frame command buffer begins recording
+        // (before any render pass has been begun); every BeginCommandRecording
+        // caller funnels through this single seam. Implemented by the renderer
+        // to prepare per-frame timer-query pools (vkCmdResetQueryPool must be
+        // recorded outside a render pass).
+        class IRecordingObserver {
+        public:
+            virtual ~IRecordingObserver() = default;
+            virtual void OnFrameCommandRecordingBegan(VkCommandBuffer commandBuffer) = 0;
+        };
+
         struct SubmitInfoPacket {
             VkPipelineStageFlags waitDstStageMask = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
             VkSemaphore waitSemaphore = VK_NULL_HANDLE;
@@ -25,7 +36,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         struct PresentInfoPacket {
             VkSemaphore waitSemaphore = VK_NULL_HANDLE;
             VkSwapchainKHR swapchain = VK_NULL_HANDLE;
-            const Uint32* imageIndex = nullptr;
+            Uint32 imageIndex = 0;
             VkPresentInfoKHR presentInfo{VK_STRUCTURE_TYPE_PRESENT_INFO_KHR};
         };
 
@@ -35,6 +46,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VkFence imageInFlightFence = VK_NULL_HANDLE;
             Bool isCommandRecording = false;
             Bool hasCommandBufferRecorded = 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;
+            // Submit-tracker index of this slot's most recent queue submission
+            // (written by the renderer at submit time).
+            Uint64 lastSubmitIndex = 0;
         };
 
         VkResult Initialize(VkDevice device, VkCommandPool commandPool, Uint32 frameCount);
@@ -53,13 +72,22 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Bool TransitionToPresent(VkImage image, VkImageLayout oldLayout,
                                  VkImageLayout presentLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR);
         SubmitInfoPacket GetSubmitInfo(Bool shouldSubmitCommandBuffer, Uint32 swapchainImageIndex) const;
-        PresentInfoPacket GetPresentInfo(VkSwapchainKHR swapchain, const Uint32& imageIndex) const;
+        PresentInfoPacket GetPresentInfo(VkSwapchainKHR swapchain, Uint32 imageIndex) const;
         VkResult WaitAndAcquireNextImage(VkDevice device, VkSwapchainKHR swapchain, Uint32& outImageIndex,
                                          Uint64 timeout = UINT64_MAX, VkFence acquireFence = VK_NULL_HANDLE);
 
+        // 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();
+
         Uint32 GetCurrentFrameIndex() const;
         Uint32 GetFrameCount() const;
 
+        // Observer may be null (no notifications). Not owned.
+        void SetRecordingObserver(IRecordingObserver* observer);
+
     private:
         void AssertValidFrameIndex(Uint32 frameIndex) const;
         void AssertValidSwapchainImageIndex(Uint32 imageIndex) const;
@@ -68,9 +96,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                            const VkSemaphoreCreateInfo& semaphoreInfo,
                                            const VkFenceCreateInfo& fenceInfo);
         void DestroySyncObjectsForFrame(VkDevice device, Uint32 frameIndex);
+        void FreeRetiredCommandBuffers(FrameData& frame);
 
         Vector<FrameData> m_frames;
         Vector<VkSemaphore> m_swapchainImageRenderFinishedSemaphores;
         Uint32 currentFrameIndex = 0;
+        IRecordingObserver* m_recordingObserver = nullptr;
+        // Stored at Initialize for retired-command-buffer management.
+        VkDevice m_device = VK_NULL_HANDLE;
+        VkCommandPool m_commandPool = VK_NULL_HANDLE;
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/PipelineFactory.cpp

@@ -8,9 +8,188 @@
 
 #include "PipelineFactory.h"
 
+
 namespace MobileGL::MG_Backend::DirectVulkan {
+    static const char* PrimitiveTopologyToString(VkPrimitiveTopology topology) {
+        switch (topology) {
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_POINT_LIST)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_LINE_LIST)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_LINE_STRIP)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_LINE_STRIP_WITH_ADJACENCY)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP_WITH_ADJACENCY)
+        ENUM_STR_CASE(VK_PRIMITIVE_TOPOLOGY_PATCH_LIST)
+        default:
+            return "VK_PRIMITIVE_TOPOLOGY_UNKNOWN";
+        }
+    }
+
+    static const char* SampleCountToString(VkSampleCountFlagBits sampleCount) {
+        switch (sampleCount) {
+        ENUM_STR_CASE(VK_SAMPLE_COUNT_1_BIT)
+        ENUM_STR_CASE(VK_SAMPLE_COUNT_2_BIT)
+        ENUM_STR_CASE(VK_SAMPLE_COUNT_4_BIT)
+        ENUM_STR_CASE(VK_SAMPLE_COUNT_8_BIT)
+        ENUM_STR_CASE(VK_SAMPLE_COUNT_16_BIT)
+        ENUM_STR_CASE(VK_SAMPLE_COUNT_32_BIT)
+        ENUM_STR_CASE(VK_SAMPLE_COUNT_64_BIT)
+        default:
+            return "VK_SAMPLE_COUNT_UNKNOWN";
+        }
+    }
+
+    static const char* CullModeToString(VkCullModeFlags cullMode) {
+        switch (cullMode) {
+        case VK_CULL_MODE_NONE:
+            return "VK_CULL_MODE_NONE";
+        case VK_CULL_MODE_FRONT_BIT:
+            return "VK_CULL_MODE_FRONT_BIT";
+        case VK_CULL_MODE_BACK_BIT:
+            return "VK_CULL_MODE_BACK_BIT";
+        case VK_CULL_MODE_FRONT_AND_BACK:
+            return "VK_CULL_MODE_FRONT_AND_BACK";
+        default:
+            return "VK_CULL_MODE_UNKNOWN";
+        }
+    }
+
+    static const char* CompareOpToString(VkCompareOp compareOp) {
+        switch (compareOp) {
+        ENUM_STR_CASE(VK_COMPARE_OP_NEVER)
+        ENUM_STR_CASE(VK_COMPARE_OP_LESS)
+        ENUM_STR_CASE(VK_COMPARE_OP_EQUAL)
+        ENUM_STR_CASE(VK_COMPARE_OP_LESS_OR_EQUAL)
+        ENUM_STR_CASE(VK_COMPARE_OP_GREATER)
+        ENUM_STR_CASE(VK_COMPARE_OP_NOT_EQUAL)
+        ENUM_STR_CASE(VK_COMPARE_OP_GREATER_OR_EQUAL)
+        ENUM_STR_CASE(VK_COMPARE_OP_ALWAYS)
+        default:
+            return "VK_COMPARE_OP_UNKNOWN";
+        }
+    }
+
+    static const char* LogicOpToString(VkLogicOp logicOp) {
+        switch (logicOp) {
+        ENUM_STR_CASE(VK_LOGIC_OP_CLEAR)
+        ENUM_STR_CASE(VK_LOGIC_OP_AND)
+        ENUM_STR_CASE(VK_LOGIC_OP_AND_REVERSE)
+        ENUM_STR_CASE(VK_LOGIC_OP_COPY)
+        ENUM_STR_CASE(VK_LOGIC_OP_AND_INVERTED)
+        ENUM_STR_CASE(VK_LOGIC_OP_NO_OP)
+        ENUM_STR_CASE(VK_LOGIC_OP_XOR)
+        ENUM_STR_CASE(VK_LOGIC_OP_OR)
+        ENUM_STR_CASE(VK_LOGIC_OP_NOR)
+        ENUM_STR_CASE(VK_LOGIC_OP_EQUIVALENT)
+        ENUM_STR_CASE(VK_LOGIC_OP_INVERT)
+        ENUM_STR_CASE(VK_LOGIC_OP_OR_REVERSE)
+        ENUM_STR_CASE(VK_LOGIC_OP_COPY_INVERTED)
+        ENUM_STR_CASE(VK_LOGIC_OP_OR_INVERTED)
+        ENUM_STR_CASE(VK_LOGIC_OP_NAND)
+        ENUM_STR_CASE(VK_LOGIC_OP_SET)
+        default:
+            return "VK_LOGIC_OP_UNKNOWN";
+        }
+    }
+
+    PipelineFactory::PipelineFactory(VkDevice device, const VulkanRendererConfig& config):
+        m_device(device), m_config(config) {
+        MOBILEGL_ASSERT(m_device != VK_NULL_HANDLE, "PipelineFactory: device is null");
+
+        if (m_config.DisablePipelineCache) {
+            MGLOG_I("DirectVulkan: pipeline cache disabled");
+            return;
+        }
+
+        VkPipelineCacheCreateInfo pipelineCacheInfo{VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO};
+        VK_VERIFY(vkCreatePipelineCache(m_device, &pipelineCacheInfo, nullptr, &m_pipelineCache),
+                  "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) {
+            vkDestroyPipelineCache(m_device, m_pipelineCache, nullptr);
+            m_pipelineCache = VK_NULL_HANDLE;
+        }
     }
 
     PipelineFactory::HashType PipelineFactory::ComputeHash(const PipelineCreatePayload& payload) const {
@@ -19,19 +198,44 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.vertexInputHash, sizeof(payload.vertexInputHash)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.pipelineLayout, sizeof(payload.pipelineLayout)));
         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.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.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.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)));
+        XXHASH_VERIFY(
+            XXH64_update(m_hashState, &payload.rasterizerDiscardEnable, sizeof(payload.rasterizerDiscardEnable)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.logicOpEnable, sizeof(payload.logicOpEnable)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.stencilTestEnable, sizeof(payload.stencilTestEnable)));
         XXHASH_VERIFY(XXH64_update(m_hashState, &payload.depthCompareOp, sizeof(payload.depthCompareOp)));
-        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.blendEnable, sizeof(payload.blendEnable)));
-        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.srcColorBlendFactor, sizeof(payload.srcColorBlendFactor)));
-        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.dstColorBlendFactor, sizeof(payload.dstColorBlendFactor)));
-        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.srcAlphaBlendFactor, sizeof(payload.srcAlphaBlendFactor)));
-        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.dstAlphaBlendFactor, sizeof(payload.dstAlphaBlendFactor)));
-        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.colorWriteMask, sizeof(payload.colorWriteMask)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.logicOp, sizeof(payload.logicOp)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.frontStencilFailOp, sizeof(payload.frontStencilFailOp)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.frontStencilPassOp, sizeof(payload.frontStencilPassOp)));
+        XXHASH_VERIFY(
+            XXH64_update(m_hashState, &payload.frontStencilDepthFailOp, sizeof(payload.frontStencilDepthFailOp)));
+        XXHASH_VERIFY(
+            XXH64_update(m_hashState, &payload.frontStencilCompareOp, sizeof(payload.frontStencilCompareOp)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.backStencilFailOp, sizeof(payload.backStencilFailOp)));
+        XXHASH_VERIFY(XXH64_update(m_hashState, &payload.backStencilPassOp, sizeof(payload.backStencilPassOp)));
+        XXHASH_VERIFY(
+            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,
+                payload.colorBlendAttachments.data(),
+                sizeof(payload.colorBlendAttachments[0]) * payload.colorAttachmentCount));
+        }
         return XXH64_digest(m_hashState);
     }
 
@@ -61,10 +265,24 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         MOBILEGL_ASSERT(payload.vertexInputState != nullptr, "PipelineFactory: vertexInputState is null");
         MOBILEGL_ASSERT(payload.pipelineLayout != VK_NULL_HANDLE, "PipelineFactory: pipelineLayout is null");
         MOBILEGL_ASSERT(payload.renderPass != VK_NULL_HANDLE, "PipelineFactory: renderPass is null");
+        MOBILEGL_ASSERT(payload.colorAttachmentCount <= PipelineCreatePayload::kMaxColorAttachments,
+                "PipelineFactory: colorAttachmentCount=%u is unexpectedly large",
+                payload.colorAttachmentCount);
+        MGLOG_D("PipelineFactory::CreatePipeline: programHash=0x%llx vertexInputHash=0x%llx colorAttachmentCount=%u subpass=%u",
+            static_cast<unsigned long long>(payload.programHash),
+            static_cast<unsigned long long>(payload.vertexInputHash),
+            payload.colorAttachmentCount,
+            payload.subpass);
 
         static constexpr VkDynamicState kDynamicStates[] = {
             VK_DYNAMIC_STATE_VIEWPORT,
-            VK_DYNAMIC_STATE_SCISSOR
+            VK_DYNAMIC_STATE_SCISSOR,
+            VK_DYNAMIC_STATE_BLEND_CONSTANTS,
+            VK_DYNAMIC_STATE_DEPTH_BIAS,
+            VK_DYNAMIC_STATE_LINE_WIDTH,
+            VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK,
+            VK_DYNAMIC_STATE_STENCIL_WRITE_MASK,
+            VK_DYNAMIC_STATE_STENCIL_REFERENCE
         };
 
         VkPipelineDynamicStateCreateInfo dynamicState{};
@@ -74,39 +292,66 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         VkPipelineInputAssemblyStateCreateInfo ia{VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO};
         ia.topology = payload.topology;
+        ia.primitiveRestartEnable = payload.primitiveRestartEnable ? VK_TRUE : VK_FALSE;
 
         VkPipelineViewportStateCreateInfo vpci{VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO};
         vpci.viewportCount = 1;
         vpci.scissorCount = 1;
 
         VkPipelineRasterizationStateCreateInfo raster{VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO};
-        raster.polygonMode = VK_POLYGON_MODE_FILL;
+        raster.polygonMode = payload.polygonMode;
         raster.cullMode = payload.cullMode;
         raster.frontFace = payload.frontFace;
+        raster.depthBiasEnable = payload.depthBiasEnable ? VK_TRUE : VK_FALSE;
+        raster.rasterizerDiscardEnable = payload.rasterizerDiscardEnable ? VK_TRUE : VK_FALSE;
         raster.lineWidth = 1.0f;
 
         VkPipelineMultisampleStateCreateInfo ms{VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO};
-        ms.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT;
+        ms.rasterizationSamples = payload.rasterizationSamples;
 
         VkPipelineDepthStencilStateCreateInfo depthStencil{VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO};
         depthStencil.depthTestEnable = payload.depthTestEnable ? VK_TRUE : VK_FALSE;
         depthStencil.depthWriteEnable = payload.depthWriteEnable ? VK_TRUE : VK_FALSE;
         depthStencil.depthCompareOp = payload.depthCompareOp;
         depthStencil.depthBoundsTestEnable = VK_FALSE;
-        depthStencil.stencilTestEnable = VK_FALSE;
+        depthStencil.stencilTestEnable = payload.stencilTestEnable ? VK_TRUE : VK_FALSE;
+        if (payload.stencilTestEnable) {
+            depthStencil.front.failOp = payload.frontStencilFailOp;
+            depthStencil.front.passOp = payload.frontStencilPassOp;
+            depthStencil.front.depthFailOp = payload.frontStencilDepthFailOp;
+            depthStencil.front.compareOp = payload.frontStencilCompareOp;
+            depthStencil.front.compareMask = 0xffffffffu;
+            depthStencil.front.writeMask = 0xffffffffu;
+            depthStencil.front.reference = 0;
+            depthStencil.back.failOp = payload.backStencilFailOp;
+            depthStencil.back.passOp = payload.backStencilPassOp;
+            depthStencil.back.depthFailOp = payload.backStencilDepthFailOp;
+            depthStencil.back.compareOp = payload.backStencilCompareOp;
+            depthStencil.back.compareMask = 0xffffffffu;
+            depthStencil.back.writeMask = 0xffffffffu;
+            depthStencil.back.reference = 0;
+        }
 
-        VkPipelineColorBlendAttachmentState colorAttach{};
-        colorAttach.colorWriteMask = payload.colorWriteMask;
-        colorAttach.blendEnable = payload.blendEnable ? VK_TRUE : VK_FALSE;
-        colorAttach.srcColorBlendFactor = payload.srcColorBlendFactor;
-        colorAttach.dstColorBlendFactor = payload.dstColorBlendFactor;
-        colorAttach.colorBlendOp = VK_BLEND_OP_ADD;
-        colorAttach.srcAlphaBlendFactor = payload.srcAlphaBlendFactor;
-        colorAttach.dstAlphaBlendFactor = payload.dstAlphaBlendFactor;
-        colorAttach.alphaBlendOp = VK_BLEND_OP_ADD;
+        Vector<VkPipelineColorBlendAttachmentState> colorAttachments(payload.colorAttachmentCount);
+        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.attachmentCount = 1;
-        blend.pAttachments = &colorAttach;
+        blend.logicOpEnable = payload.logicOpEnable ? VK_TRUE : VK_FALSE;
+        blend.logicOp = payload.logicOp;
+        blend.attachmentCount = payload.colorAttachmentCount;
+        blend.pAttachments = colorAttachments.empty() ? nullptr : colorAttachments.data();
 
         VkGraphicsPipelineCreateInfo gpi{VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO};
         gpi.stageCount = static_cast<Uint32>(payload.stages->size());
@@ -124,8 +369,52 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         gpi.subpass = payload.subpass;
 
         VkPipeline pipeline = VK_NULL_HANDLE;
-        VK_VERIFY(vkCreateGraphicsPipelines(m_device, VK_NULL_HANDLE, 1, &gpi, nullptr, &pipeline),
-                  "vkCreateGraphicsPipelines");
+        const VkResult result = vkCreateGraphicsPipelines(m_device, m_pipelineCache, 1, &gpi, nullptr, &pipeline);
+        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),
+                    result,
+                    static_cast<unsigned long long>(payload.programHash),
+                    static_cast<unsigned long long>(payload.vertexInputHash),
+                    gpi.stageCount,
+                    PrimitiveTopologyToString(payload.topology),
+                    payload.topology,
+                    payload.colorAttachmentCount,
+                    SampleCountToString(payload.rasterizationSamples),
+                    payload.rasterizationSamples,
+                    payload.subpass);
+            MGLOG_F("PipelineFactory::CreatePipeline state: cullMode=%s(0x%x) frontFace=%d depthTest=%d depthWrite=%d depthCompare=%s(%d) depthBias=%d rasterizerDiscard=%d stencilTest=%d logicOpEnable=%d logicOp=%s(%d)",
+                    CullModeToString(payload.cullMode),
+                    static_cast<Uint32>(payload.cullMode),
+                    payload.frontFace,
+                    payload.depthTestEnable ? 1 : 0,
+                    payload.depthWriteEnable ? 1 : 0,
+                    CompareOpToString(payload.depthCompareOp),
+                    payload.depthCompareOp,
+                    payload.depthBiasEnable ? 1 : 0,
+                    payload.rasterizerDiscardEnable ? 1 : 0,
+                    payload.stencilTestEnable ? 1 : 0,
+                    payload.logicOpEnable ? 1 : 0,
+                    LogicOpToString(payload.logicOp),
+                    payload.logicOp);
+            MGLOG_F("PipelineFactory::CreatePipeline vertex input: bindingCount=%u attributeCount=%u",
+                    payload.vertexInputState->vertexBindingDescriptionCount,
+                    payload.vertexInputState->vertexAttributeDescriptionCount);
+            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",
+                        i,
+                        attachment.blendEnable == VK_TRUE ? 1 : 0,
+                        static_cast<Uint32>(attachment.colorWriteMask),
+                        attachment.srcColorBlendFactor,
+                        attachment.dstColorBlendFactor,
+                        attachment.colorBlendOp,
+                        attachment.srcAlphaBlendFactor,
+                        attachment.dstAlphaBlendFactor,
+                        attachment.alphaBlendOp);
+            }
+        }
+        VK_VERIFY(result, "vkCreateGraphicsPipelines");
         return pipeline;
     }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/PipelineFactory.h

@@ -10,6 +10,7 @@
 
 #include "Config.h"
 #include "../VkIncludes.h"
+#include "MG_State/GLState/FramebufferState/FramebufferObject.h"
 #include <Includes.h>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
@@ -18,31 +19,45 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         using HashType = Uint64;
 
         struct PipelineCreatePayload {
+            static constexpr Uint32 kMaxColorAttachments = MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS;
+
             HashType programHash = 0;
             HashType vertexInputHash = 0;
             VkPipelineLayout pipelineLayout = VK_NULL_HANDLE;
             VkRenderPass renderPass = VK_NULL_HANDLE;
+            Uint32 colorAttachmentCount = 1;
+            VkSampleCountFlagBits rasterizationSamples = VK_SAMPLE_COUNT_1_BIT;
             Uint32 subpass = 0;
             VkPrimitiveTopology topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST;
+            Bool primitiveRestartEnable = false;
+            VkPolygonMode polygonMode = VK_POLYGON_MODE_FILL;
             VkCullModeFlags cullMode = VK_CULL_MODE_BACK_BIT;
             VkFrontFace frontFace = VK_FRONT_FACE_CLOCKWISE;
             Bool depthTestEnable = false;
             Bool depthWriteEnable = false;
+            Bool depthBiasEnable = false;
+            Bool rasterizerDiscardEnable = false;
+            Bool logicOpEnable = false;
+            Bool stencilTestEnable = false;
             VkCompareOp depthCompareOp = VK_COMPARE_OP_ALWAYS;
-            Bool blendEnable = false;
-            VkBlendFactor srcColorBlendFactor = VK_BLEND_FACTOR_ONE;
-            VkBlendFactor dstColorBlendFactor = VK_BLEND_FACTOR_ZERO;
-            VkBlendFactor srcAlphaBlendFactor = VK_BLEND_FACTOR_ONE;
-            VkBlendFactor dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO;
-            VkColorComponentFlags colorWriteMask =
-                VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT |
-                VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT;
+            VkLogicOp logicOp = VK_LOGIC_OP_COPY;
+            VkStencilOp frontStencilFailOp = VK_STENCIL_OP_KEEP;
+            VkStencilOp frontStencilPassOp = VK_STENCIL_OP_KEEP;
+            VkStencilOp frontStencilDepthFailOp = VK_STENCIL_OP_KEEP;
+            VkCompareOp frontStencilCompareOp = VK_COMPARE_OP_ALWAYS;
+            VkStencilOp backStencilFailOp = VK_STENCIL_OP_KEEP;
+            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;
             const VkPipelineVertexInputStateCreateInfo* vertexInputState = nullptr;
         };
 
-        explicit PipelineFactory(VkDevice device, const VulkanRendererConfig& config):
-            m_device(device), m_config(config) {}
+        explicit PipelineFactory(VkDevice device, const VulkanRendererConfig& config);
         ~PipelineFactory();
         PipelineFactory(const PipelineFactory&) = delete;
 
@@ -50,12 +65,35 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkPipeline GetOrCreatePipeline(const PipelineCreatePayload& payload);
         void DestroyAll();
 
+        // 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:
         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;
         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

@@ -14,14 +14,25 @@
 #include "MG_State/GLState/TextureState/TextureEnum.h"
 
 #include <Includes.h>
+#include <spirv_reflect.h>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
+    enum class SamplerNumericDomain : Uint8 {
+        Unknown = 0,
+        Float,
+        SignedInteger,
+        UnsignedInteger,
+    };
+
     class ProgramFactory {
     public:
         enum class DescriptorBindingKind : Uint8 {
             None = 0,
             UniformBufferDynamic,
-            CombinedImageSampler
+            CombinedImageSampler,
+            UniformTexelBuffer,
+            StorageBuffer,
+            StorageImage
         };
 
         enum class CompileOptionBit : Uint {
@@ -36,6 +47,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         using HashType = Uint64;
 
         struct VkProgramObject {
+            static constexpr Uint32 kMaxVertexInputLocations = 32;
+
             HashType hash = 0;
             Vector<VkPipelineShaderStageCreateInfo> stages;
             Vector<VkShaderModule> modules;
@@ -45,9 +58,36 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             VkPipelineLayout pipelineLayout = VK_NULL_HANDLE;
             Vector<DescriptorBindingKind> bindingKinds;
             Vector<Uint32> dynamicBindings;
+            Vector<Int> uniformBlockIndexByBinding;
+            // Descriptor count per binding (1 except for UBO instance arrays, which occupy 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;
+            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;
             Int globalUboBinding = -1;
+            Uint32 activeVertexInputLocationMask = 0;
+            Array<GLenum, kMaxVertexInputLocations> vertexInputTypes{};
+            Uint32 activeFragmentOutputLocationMask = 0;
+            Array<GLenum, kMaxVertexInputLocations> fragmentOutputTypes{};
+            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;
 
             static inline VkDevice s_device = VK_NULL_HANDLE;
 
@@ -62,13 +102,39 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 pipelineLayout = other.pipelineLayout;
                 bindingKinds = std::move(other.bindingKinds);
                 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);
+                samplerNumericDomainByBinding = std::move(other.samplerNumericDomainByBinding);
+                storageImageFormatByBinding = std::move(other.storageImageFormatByBinding);
+                storageImageUsesBindingFormatByBinding =
+                    std::move(other.storageImageUsesBindingFormatByBinding);
+                storageBlockNameByBinding = std::move(other.storageBlockNameByBinding);
+                storageBlockIndexByBinding = std::move(other.storageBlockIndexByBinding);
+                hasStorageImages = other.hasStorageImages;
                 globalUboBinding = other.globalUboBinding;
+                activeVertexInputLocationMask = other.activeVertexInputLocationMask;
+                vertexInputTypes = other.vertexInputTypes;
+                activeFragmentOutputLocationMask = other.activeFragmentOutputLocationMask;
+                fragmentOutputTypes = other.fragmentOutputTypes;
+                rasterizationProducerStage = other.rasterizationProducerStage;
+                producerOutputComponentCount = other.producerOutputComponentCount;
+                fragmentInputComponentCount = other.fragmentInputComponentCount;
+                fragmentReplacesDepth = other.fragmentReplacesDepth;
                 other.hash = 0;
                 other.descriptorSetLayout = VK_NULL_HANDLE;
                 other.pipelineLayout = VK_NULL_HANDLE;
+                other.hasStorageImages = false;
                 other.globalUboBinding = -1;
+                other.activeVertexInputLocationMask = 0;
+                other.activeFragmentOutputLocationMask = 0;
+                other.rasterizationProducerStage = ShaderStage::Unknown;
+                other.producerOutputComponentCount = 0;
+                other.fragmentInputComponentCount = 0;
+                other.fragmentReplacesDepth = false;
             }
             VkProgramObject& operator=(VkProgramObject&& other) noexcept {
                 if (this == &other) {
@@ -82,13 +148,39 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                 pipelineLayout = other.pipelineLayout;
                 bindingKinds = std::move(other.bindingKinds);
                 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);
+                samplerNumericDomainByBinding = std::move(other.samplerNumericDomainByBinding);
+                storageImageFormatByBinding = std::move(other.storageImageFormatByBinding);
+                storageImageUsesBindingFormatByBinding =
+                    std::move(other.storageImageUsesBindingFormatByBinding);
+                storageBlockNameByBinding = std::move(other.storageBlockNameByBinding);
+                storageBlockIndexByBinding = std::move(other.storageBlockIndexByBinding);
+                hasStorageImages = other.hasStorageImages;
                 globalUboBinding = other.globalUboBinding;
+                activeVertexInputLocationMask = other.activeVertexInputLocationMask;
+                vertexInputTypes = other.vertexInputTypes;
+                activeFragmentOutputLocationMask = other.activeFragmentOutputLocationMask;
+                fragmentOutputTypes = other.fragmentOutputTypes;
+                rasterizationProducerStage = other.rasterizationProducerStage;
+                producerOutputComponentCount = other.producerOutputComponentCount;
+                fragmentInputComponentCount = other.fragmentInputComponentCount;
+                fragmentReplacesDepth = other.fragmentReplacesDepth;
                 other.hash = 0;
                 other.descriptorSetLayout = VK_NULL_HANDLE;
                 other.pipelineLayout = VK_NULL_HANDLE;
+                other.hasStorageImages = false;
                 other.globalUboBinding = -1;
+                other.activeVertexInputLocationMask = 0;
+                other.activeFragmentOutputLocationMask = 0;
+                other.rasterizationProducerStage = ShaderStage::Unknown;
+                other.producerOutputComponentCount = 0;
+                other.fragmentInputComponentCount = 0;
+                other.fragmentReplacesDepth = false;
                 return *this;
             }
 
@@ -118,8 +210,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             }
         };
 
-        explicit ProgramFactory(VkDevice device, const VulkanRendererConfig& config, Uint32 maxBindings = 16)
-            : m_device(device), m_config(config), m_maxBindings(maxBindings) {
+        explicit ProgramFactory(VkDevice device, const VulkanRendererConfig& config, Uint32 maxBindings = 16,
+                                Bool shaderDrawParametersEnabled = false,
+                                Bool unformattedFloatStorageImagesEnabled = false)
+            : m_device(device), m_maxBindings(maxBindings), m_config(config),
+              m_shaderDrawParametersEnabled(shaderDrawParametersEnabled),
+              m_unformattedFloatStorageImagesEnabled(unformattedFloatStorageImagesEnabled) {
             VkProgramObject::s_device = device;
         }
         ~ProgramFactory() = default;
@@ -130,15 +226,46 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             const MG_State::GLState::ProgramObject& program, CompileOptionFlags flags);
 
         static VkShaderStageFlagBits ToVkStage(ShaderStage stage);
+        static VkFormat ConvertSpirvImageFormatToVkFormat(SpvImageFormat format);
+        static SamplerNumericDomain UniformTypeToSamplerNumericDomain(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);
 
     private:
+        struct ProgramLookupCache {
+            const MG_State::GLState::ProgramObject* program = nullptr;
+            Uint32 backendStateVersion = 0;
+            CompileOptionFlags flags{};
+            HashType hash = 0;
+        };
+
         static TextureTarget UniformTypeToTextureTarget(GLenum glType);
-        void ReflectLayout(const MG_State::GLState::ProgramObject& program, VkProgramObject& entry) const;
+        void ReflectVertexInputs(const Vector<SharedPtr<MG_State::GLState::ShaderObject>>& shaders,
+                     const Vector<Vector<Uint>>& spirv,
+                     VkProgramObject& entry) const;
+        void ReflectFragmentOutputs(const Vector<SharedPtr<MG_State::GLState::ShaderObject>>& shaders,
+                        const Vector<Vector<Uint>>& spirv,
+                        VkProgramObject& entry) const;
+        void ReflectLayout(const MG_State::GLState::ProgramObject& program, const Vector<Vector<Uint>>& spirv,
+                           VkProgramObject& entry) const;
 
         VkDevice m_device = VK_NULL_HANDLE;
         Uint32 m_maxBindings = 0;
         UnorderedMap<HashType, VkProgramObject> m_cache;
         const VulkanRendererConfig& m_config;
+        // True when the device enabled shaderDrawParameters; gates the InstanceIndex rebase pass
+        // (which needs the DrawParameters capability / gl_BaseInstance builtin).
+        Bool m_shaderDrawParametersEnabled = false;
+        // True only when the logical device enabled both
+        // shaderStorageImageReadWithoutFormat and shaderStorageImageWriteWithoutFormat.
+        Bool m_unformattedFloatStorageImagesEnabled = false;
+        mutable ProgramLookupCache m_lastLookup;
         static inline XXH64_state_t* m_hashState = XXH64_createState();
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/SwapchainObject.cpp

@@ -31,8 +31,19 @@ static const char* string_VkColorSpaceKHR(VkColorSpaceKHR) {
     return "VkColorSpaceKHR(unknown)";
 }
 
-static const char* string_VkPresentModeKHR(VkPresentModeKHR) {
-    return "VkPresentModeKHR(unknown)";
+static const char* string_VkPresentModeKHR(VkPresentModeKHR presentMode) {
+    switch (presentMode) {
+    case VK_PRESENT_MODE_IMMEDIATE_KHR:
+        return "VK_PRESENT_MODE_IMMEDIATE_KHR";
+    case VK_PRESENT_MODE_MAILBOX_KHR:
+        return "VK_PRESENT_MODE_MAILBOX_KHR";
+    case VK_PRESENT_MODE_FIFO_KHR:
+        return "VK_PRESENT_MODE_FIFO_KHR";
+    case VK_PRESENT_MODE_FIFO_RELAXED_KHR:
+        return "VK_PRESENT_MODE_FIFO_RELAXED_KHR";
+    default:
+        return "VkPresentModeKHR(unknown)";
+    }
 }
 
 static const char* string_VkSurfaceTransformFlagBitsKHR(VkSurfaceTransformFlagBitsKHR) {
@@ -105,7 +116,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     VkSurfaceFormatKHR SwapchainObject::ChooseSwapchainSurfaceFormat(
         const Vector<VkSurfaceFormatKHR>& availableFormats) {
         for (const auto& availableFormat : availableFormats) {
-            if (availableFormat.format == VK_FORMAT_B8G8R8A8_SRGB &&
+            if ((availableFormat.format == VK_FORMAT_B8G8R8A8_UNORM ||
+                 availableFormat.format == VK_FORMAT_R8G8B8A8_UNORM) &&
+                availableFormat.colorSpace == VK_COLOR_SPACE_SRGB_NONLINEAR_KHR) {
+                return availableFormat;
+            }
+        }
+        for (const auto& availableFormat : availableFormats) {
+            if ((availableFormat.format == VK_FORMAT_B8G8R8A8_SRGB ||
+                 availableFormat.format == VK_FORMAT_R8G8B8A8_SRGB) &&
                 availableFormat.colorSpace == VK_COLOR_SPACE_SRGB_NONLINEAR_KHR) {
                 return availableFormat;
             }
@@ -130,7 +149,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     }
 
     void SwapchainObject::Create(VkDevice device, VkPhysicalDevice physicalDevice, VkSurfaceKHR surface,
-                                 Uint32 graphicsQueueFamily, Uint32 presentQueueFamily, Uint32 minImageCountHint) {
+                                 Uint32 graphicsQueueFamily, Uint32 presentQueueFamily, Uint32 minImageCountHint,
+                                 VkExtent2D desiredExtent) {
         const auto swapchainCapabilities = GetSwapchainCapabilities(physicalDevice, surface);
         MOBILEGL_ASSERT(swapchainCapabilities.IsComplete(),
                         "SwapchainObject::Create failed: incomplete swapchain capabilities");
@@ -153,7 +173,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         MGLOG_I("Picked present mode: %s", string_VkPresentModeKHR(presentMode));
 
         const auto& swapchainCaps = swapchainCapabilities.capabilities;
-        const auto targetImageCount = std::max<Uint32>(minImageCountHint, swapchainCaps.minImageCount);
+        Uint32 targetImageCount = std::max<Uint32>(minImageCountHint, swapchainCaps.minImageCount);
+        if (swapchainCaps.maxImageCount != 0) {
+            targetImageCount = std::min(targetImageCount, swapchainCaps.maxImageCount);
+        }
         MGLOG_I("Set minImageCount = %u", targetImageCount);
         MGLOG_I("Swapchain currentTransform = %s",
                 string_VkSurfaceTransformFlagBitsKHR(swapchainCaps.currentTransform));
@@ -164,6 +187,15 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         createInfo.imageFormat = pickedSurfaceFormat.format;
         createInfo.imageColorSpace = pickedSurfaceFormat.colorSpace;
         createInfo.imageExtent = swapchainCaps.currentExtent;
+        if (createInfo.imageExtent.width == UINT32_MAX || createInfo.imageExtent.height == UINT32_MAX) {
+            createInfo.imageExtent.width = std::clamp(desiredExtent.width,
+                                                      swapchainCaps.minImageExtent.width,
+                                                      swapchainCaps.maxImageExtent.width);
+            createInfo.imageExtent.height = std::clamp(desiredExtent.height,
+                                                       swapchainCaps.minImageExtent.height,
+                                                       swapchainCaps.maxImageExtent.height);
+        }
+        const VkExtent2D defaultFramebufferExtent = createInfo.imageExtent;
         if (swapchainCaps.currentTransform == VK_SURFACE_TRANSFORM_ROTATE_90_BIT_KHR ||
             swapchainCaps.currentTransform == VK_SURFACE_TRANSFORM_ROTATE_270_BIT_KHR) {
             std::swap(createInfo.imageExtent.width, createInfo.imageExtent.height);
@@ -234,11 +266,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         // Properly initialize Default FBO here
         auto& defaultFBOInfo = MG_Impl::GLImpl::FramebufferImpl::pDefaultFramebufferInfo;
+        const Int extentWidth = static_cast<Int>(defaultFramebufferExtent.width);
+        const Int extentHeight = static_cast<Int>(defaultFramebufferExtent.height);
+        const SizeT defaultAttachmentByteSize =
+            static_cast<SizeT>(defaultFramebufferExtent.width) *
+            static_cast<SizeT>(defaultFramebufferExtent.height) * 4;
+
         auto* colorTex = static_cast<MG_State::GLState::TextureObject2D*>(defaultFBOInfo->colorAttachment.get());
         colorTex->AllocateStorage(
             TextureUploadTarget::Texture2D, 0, {
-                {(Int)createInfo.imageExtent.width, (Int)createInfo.imageExtent.height, 1},
-                createInfo.imageExtent.width * (Int)createInfo.imageExtent.height * 4}); // TODO: 4 is format size
+                {extentWidth, extentHeight, 1},
+                defaultAttachmentByteSize}); // TODO: 4 is format size
         TextureInternalFormat depthFormat = TextureInternalFormat::Depth24Stencil8;
         switch (m_depthStencilFormat) {
             case VK_FORMAT_D24_UNORM_S8_UINT:
@@ -257,8 +295,32 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         auto* depthTex = static_cast<MG_State::GLState::TextureObject2D*>(defaultFBOInfo->depthAttachment.get());
         depthTex->SetInternalFormat(depthFormat);
         depthTex->AllocateStorage(TextureUploadTarget::Texture2D, 0, {
-            {(Int)createInfo.imageExtent.width, (Int)createInfo.imageExtent.height, 1},
-                        createInfo.imageExtent.width * createInfo.imageExtent.width * 4}); // TODO: 4 is format size
+            {extentWidth, extentHeight, 1},
+            defaultAttachmentByteSize}); // TODO: 4 is format size
+
+        // The default FBO's stencil attachment must track the swapchain extent:
+        // FramebufferObject::CheckCompleteness requires every valid attachment
+        // to share the same dimensions, and Init.cpp leaves a 512x512 placeholder.
+        // Without this the retrace-layer glReadPixels snapshot fails with
+        // GL_INVALID_FRAMEBUFFER_OPERATION on DirectVulkan.
+        TextureInternalFormat stencilFormat = TextureInternalFormat::Depth24Stencil8;
+        switch (m_depthStencilFormat) {
+            case VK_FORMAT_D32_SFLOAT_S8_UINT:
+                stencilFormat = TextureInternalFormat::Depth32FStencil8;
+                break;
+            case VK_FORMAT_D24_UNORM_S8_UINT:
+                stencilFormat = TextureInternalFormat::Depth24Stencil8;
+                break;
+            default:
+                // No stencil plane; mirror the depth format for consistency.
+                stencilFormat = depthFormat;
+                break;
+        }
+        auto* stencilTex = static_cast<MG_State::GLState::TextureObject2D*>(defaultFBOInfo->stencilAttachment.get());
+        stencilTex->SetInternalFormat(stencilFormat);
+        stencilTex->AllocateStorage(TextureUploadTarget::Texture2D, 0, {
+            {extentWidth, extentHeight, 1},
+            defaultAttachmentByteSize}); // TODO: 4 is format size
 
     }
 
@@ -290,7 +352,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             imageInfo.format = m_depthStencilFormat;
             imageInfo.tiling = VK_IMAGE_TILING_OPTIMAL;
             imageInfo.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
-            imageInfo.usage = VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
+            imageInfo.usage = VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
             imageInfo.samples = VK_SAMPLE_COUNT_1_BIT;
             imageInfo.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
             VK_VERIFY(vkCreateImage(device, &imageInfo, nullptr, &m_depthStencilImages[i]), "vkCreateImage(depth)");

MobileGL/MG_Backend/DirectVulkan/Renderer/SwapchainObject.h

@@ -29,7 +29,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         static VkPresentModeKHR ChooseSwapchainPresentMode(const Vector<VkPresentModeKHR>& availablePresentModes);
 
         void Create(VkDevice device, VkPhysicalDevice physicalDevice, VkSurfaceKHR surface, Uint32 graphicsQueueFamily,
-                    Uint32 presentQueueFamily, Uint32 minImageCountHint);
+                    Uint32 presentQueueFamily, Uint32 minImageCountHint, VkExtent2D desiredExtent);
         void Shutdown(VkDevice device);
 
         VkSwapchainKHR GetHandle() const { return m_swapchain; }

MobileGL/MG_Backend/DirectVulkan/Renderer/UniformManager.h

@@ -28,6 +28,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Uint32 binding = 0;
             MG_State::GLState::ITextureObject* texture = nullptr;
             const MG_State::GLState::SamplerObject* sampler = nullptr;
+            VkImageView imageView = VK_NULL_HANDLE;
         };
 
         Bool Initialize(VkDevice device, VkBufferManager* bufferManager,
@@ -41,12 +42,22 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Bool CollectSampledTextures(const MG_State::GLState::ProgramObject& program,
                                     const ProgramFactory::VkProgramObject& programObj,
                                     Vector<MG_State::GLState::ITextureObject*>& outTextures);
+        Bool CollectStorageImageTextures(const MG_State::GLState::ProgramObject& program,
+                                         const ProgramFactory::VkProgramObject& programObj,
+                                         Vector<MG_State::GLState::ITextureObject*>& outTextures) const;
         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);
 
+        // Pure format-policy helper kept public for host regression tests. Formatted storage
+        // images use their shader qualifier; transformed float images use glBindImageTexture's
+        // format and never silently fall back to the backing image format.
+        static VkFormat ResolveStorageImageViewFormat(VkFormat reflectedFormat, GLenum bindingFormat,
+                                                      VkFormat resourceFormat, Bool useBindingFormat);
+
     private:
         struct DescriptorPoolBucket {
             VkDescriptorPool handle = VK_NULL_HANDLE;
@@ -54,27 +65,65 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Uint32 allocatedSets = 0;
         };
 
+        struct DescriptorSetCacheEntry {
+            Vector<VkDescriptorSet> sets;
+            Uint32 cursor = 0;
+        };
+
         struct FrameResources {
             Vector<DescriptorPoolBucket> descriptorPools;
+            UnorderedMap<VkDescriptorSetLayout, DescriptorSetCacheEntry> descriptorSetCacheByLayout;
+            Vector<VkBufferView> texelBufferViews;
             Uint32 activeDescriptorPoolIndex = 0;
             Uint32 allocatedSetsThisFrame = 0;
             Uint32 peakAllocatedSetsThisFrame = 0;
         };
 
-        Bool ResolveSamplerTexture(const MG_State::GLState::ProgramObject& program,
+        static Bool ResolveSamplerTexture(const MG_State::GLState::ProgramObject& program,
                                    const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
-                                   SharedPtr<MG_State::GLState::ITextureObject>& outTexture) const;
+                                   SharedPtr<MG_State::GLState::ITextureObject>& outTexture);
+        // 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;
         Bool ResolveSamplerDescriptor(VkCommandBuffer commandBuffer, const MG_State::GLState::ProgramObject& program,
                                       const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
                                       VkDescriptorImageInfo& outImageInfo) const;
         Bool ResolveSamplerDescriptorOverride(const SamplerBindingOverride& samplerBindingOverride,
                                               VkDescriptorImageInfo& outImageInfo) const;
-        Bool GatherBindingPayloads(const MG_State::GLState::ProgramObject& program, Vector<const void*>& outData,
-                                   Vector<VkDeviceSize>& outSizes) const;
+        Bool ResolveTexelBufferDescriptor(const MG_State::GLState::ProgramObject& program,
+                                          const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
+                                          Uint32 frameIndex, VkBufferView& outBufferView);
+        Bool ResolveStorageBufferDescriptor(const MG_State::GLState::ProgramObject& program,
+                                            const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
+                                            VkDescriptorBufferInfo& outBufferInfo) const;
+        Bool ResolveStorageImageDescriptor(VkCommandBuffer commandBuffer,
+                                           const MG_State::GLState::ProgramObject& program,
+                                           const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
+                                           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 {
+            Bool directBindable = false;
+            VkBuffer buffer = VK_NULL_HANDLE;
+            VkDeviceSize range = 0;         // reflected block size; constant across draws (hashed)
+            VkDeviceSize dynamicOffset = 0; // block range start; moves per draw (NOT hashed)
+            const void* payload = nullptr;  // fallback UploadTransient path
+            VkDeviceSize payloadSize = 0;
+        };
+        Bool ResolveUniformBufferPayload(const MG_State::GLState::ProgramObject& program,
+                                         const ProgramFactory::VkProgramObject& programObj, Uint32 binding,
+                                         Uint32 arrayElement, UboBindResult& out) const;
         Bool CreateDescriptorPool(Uint32 maxSets, VkDescriptorPool& outPool) const;
         Bool GrowFrameDescriptorPool(FrameResources& frame, Uint32 frameIndex);
         VkResult AllocateDescriptorSetsFromActivePool(
             Uint32 frameIndex, const ProgramFactory::VkProgramObject& programObj, VkDescriptorSet& outDescriptorSet);
+        VkResult AcquireDescriptorSet(Uint32 frameIndex,
+                                      const ProgramFactory::VkProgramObject& programObj,
+                                      VkDescriptorSet& outDescriptorSet);
 
         VkDevice m_device = VK_NULL_HANDLE;
         VkBufferManager* m_bufferManager = nullptr;
@@ -88,6 +137,53 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Uint32 m_peakDescriptorSetsObserved = 0;
         VkTextureManager* m_textureManager = nullptr;
         VkSamplerManager* m_samplerManager = nullptr;
+        mutable SharedPtr<MG_State::GLState::ITextureObject> m_fallbackTexture2D;
+
+        // Per-draw scratch buffers for BindProgramUniformBuffers: reused (clear keeps
+        // capacity) so the descriptor-write path stops allocating on every draw.
+        Vector<VkWriteDescriptorSet> m_writesScratch;
+        Vector<VkDescriptorBufferInfo> m_bufferInfosScratch;
+        Vector<VkDescriptorImageInfo> m_imageInfosScratch;
+        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;
+
+        // 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
+        // sampler objects with identical state still resolve to one VkSampler. This memo only
+        // skips recomputing that hash. Across a draw batch the bound sampler set is stable, so a
+        // binding whose sampler (lifetime id + version, bumped on every setter) and texture
+        // (lifetime id + params version, bumped on the format/border-color setters that feed the
+        // key) are unchanged recycles the VkSampler it resolved last draw; a param change bumps
+        // a version and forces a re-resolve. Both objects are keyed by a never-reused monotonic
+        // 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.
+        struct SamplerResolveMemo {
+            Uint64 samplerLifetimeId = 0;
+            Uint64 textureLifetimeId = 0;
+            VkSampler sampler = VK_NULL_HANDLE;
+            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;
+        };
+        mutable Vector<SamplerResolveMemo> m_samplerResolveMemo;
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan
-

MobileGL/MG_Backend/DirectVulkan/Renderer/VertexInputStateFactory.cpp

@@ -29,6 +29,7 @@ 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.IsBgra, sizeof(attr.IsBgra)));
             XXHASH_VERIFY(XXH64_update(m_hashState, &attr.Divisor, sizeof(attr.Divisor)));
 
             const SizeT bufferKey = reinterpret_cast<SizeT>(attr.Buffer.get());
@@ -38,70 +39,127 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         return XXH64_digest(m_hashState);
     }
 
+    VertexInputStateFactory::HashType VertexInputStateFactory::GetOrComputeHash(
+        const MG_State::GLState::VertexArrayObject& vao) const {
+        HashType hash = 0;
+        if (!vao.GetBackendHashMemo(hash)) {
+            hash = ComputeHash(vao);
+            vao.SetBackendHashMemo(hash);
+        }
+        return hash;
+    }
+
     const VertexInputStateFactory::BackendVertexInputState& VertexInputStateFactory::GetOrCreateVertexInputState(
         const MG_State::GLState::VertexArrayObject& vao) {
-        const HashType hash = ComputeHash(vao);
+        return GetOrCreateVertexInputState(vao, GetOrComputeHash(vao));
+    }
+
+    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;
         }
 
         VertexInputStateBuilder builder;
-        UnorderedMap<SizeT, Uint32> bindingByBufferKey;
-        UnorderedMap<SizeT, Uint32> strideByBufferKey;
-        UnorderedMap<SizeT, VkVertexInputRate> inputRateByBufferKey;
         Vector<SizeT> bindingBufferKeys;
+        Vector<SizeT> bindingBaseOffsets;
+        Vector<Uint32> bindingAttributeLocations;
+        Vector<Bool> bindingUsesClientMemory;
+        Vector<VertexStreamConversion> bindingConversions;
+        Uint32 unsupportedAttribMask = 0;
 
         for (Uint32 location = 0; location < MG_State::GLState::VertexArrayObject::MAX_VERTEX_ATTRIBS; ++location) {
             const auto& attr = vao.GetAttribute(location);
-            if (!attr.Enabled || !attr.Buffer) {
+            if (!attr.Enabled) {
                 continue;
             }
 
-            const auto vkFormat = ToVkVertexFormat(attr.Type, attr.Size, attr.Normalized, attr.IsInteger);
-            if (vkFormat == VK_FORMAT_UNDEFINED) {
-                MGLOG_D("Skipping unsupported vertex attribute layout (location=%u, type=%s, size=%d)",
+            const VkFormat sourceVkFormat =
+                ToVkVertexFormat(attr.Type, attr.Size, attr.Normalized, attr.IsInteger, attr.IsBgra);
+            if (sourceVkFormat == VK_FORMAT_UNDEFINED) {
+                MGLOG_E("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);
                 continue;
             }
 
-            const SizeT componentSize = GetComponentSize(attr.Type);
-            if (componentSize == 0) {
-                MGLOG_D("Skipping vertex attribute with unknown component size (location=%u, type=%s)",
+            VkFormat vkFormat = sourceVkFormat;
+            VertexStreamConversion conversion = VertexStreamConversion::None;
+            if (!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 "
+                                "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),
+                                MG_Util::ConvertDataTypeToString(attr.Type).c_str(), attr.Size,
+                                attr.Normalized ? "true" : "false", attr.IsInteger ? "true" : "false");
+                    }
+                }
+
+                if (conversion == VertexStreamConversion::None) {
+                    MGLOG_E("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);
+                    unsupportedAttribMask |= (1u << location);
+                    continue;
+                }
+            }
+
+            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 "
+                        "enabled but cannot be sized",
                         location, MG_Util::ConvertDataTypeToString(attr.Type).c_str());
+                unsupportedAttribMask |= (1u << location);
                 continue;
             }
 
-            const Uint32 stride = attr.Stride > 0
-                                      ? static_cast<Uint32>(attr.Stride)
-                                      : static_cast<Uint32>(componentSize * static_cast<SizeT>(attr.Size));
+            const Uint32 sourceStride =
+                attr.Stride > 0 ? static_cast<Uint32>(attr.Stride) : static_cast<Uint32>(attribByteSize);
+            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 ||
+                 (!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 "
+                        "(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)));
+            }
             const VkVertexInputRate inputRate =
                 (attr.Divisor == 0) ? VK_VERTEX_INPUT_RATE_VERTEX : VK_VERTEX_INPUT_RATE_INSTANCE;
 
             const SizeT bufferKey = reinterpret_cast<SizeT>(attr.Buffer.get());
-            Uint32 binding = 0;
-            auto itBinding = bindingByBufferKey.find(bufferKey);
-            if (itBinding == bindingByBufferKey.end()) {
-                binding = static_cast<Uint32>(bindingByBufferKey.size());
-                bindingByBufferKey.emplace(bufferKey, binding);
-                strideByBufferKey.emplace(bufferKey, stride);
-                inputRateByBufferKey.emplace(bufferKey, inputRate);
-                bindingBufferKeys.push_back(bufferKey);
-                builder.AddBinding(binding, stride, inputRate);
-            } else {
-                binding = itBinding->second;
-                if (strideByBufferKey[bufferKey] != stride) {
-                    MGLOG_D("Skipping vertex attribute at location %u: stride mismatch (%u vs %u) on same buffer",
-                            location, stride, strideByBufferKey[bufferKey]);
-                    continue;
-                }
-                if (inputRateByBufferKey[bufferKey] != inputRate) {
-                    MGLOG_D("Skipping vertex attribute at location %u: input-rate mismatch on same buffer", location);
-                    continue;
-                }
-            }
-
-            builder.AddAttribute(location, binding, vkFormat, static_cast<Uint32>(attr.Offset));
+            const Uint32 binding = static_cast<Uint32>(bindingBufferKeys.size());
+            bindingBufferKeys.push_back(bufferKey);
+            bindingBaseOffsets.push_back(attr.Buffer ? attr.Offset : 0);
+            bindingAttributeLocations.push_back(location);
+            bindingUsesClientMemory.push_back(attr.Buffer == nullptr);
+            bindingConversions.push_back(conversion);
+            builder.AddBinding(binding, stride, inputRate);
+            builder.AddAttribute(location, binding, vkFormat, 0);
         }
 
         const auto& state = builder.Build();
@@ -111,14 +169,43 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         entry.bindings = builder.GetBindings();
         entry.attributes = builder.GetAttributes();
         entry.bindingBufferKeys = std::move(bindingBufferKeys);
+        entry.bindingBaseOffsets = std::move(bindingBaseOffsets);
+        entry.bindingAttributeLocations = std::move(bindingAttributeLocations);
+        entry.bindingUsesClientMemory = std::move(bindingUsesClientMemory);
+        entry.bindingConversions = std::move(bindingConversions);
+        entry.unsupportedAttribMask = unsupportedAttribMask;
         entry.state = state;
         entry.state.pVertexBindingDescriptions = entry.bindings.empty() ? nullptr : entry.bindings.data();
         entry.state.pVertexAttributeDescriptions = entry.attributes.empty() ? nullptr : entry.attributes.data();
         return entry;
     }
 
-    VkFormat VertexInputStateFactory::ToVkVertexFormat(DataType type, Int size, Bool normalized, Bool isInteger) {
+    VkFormat VertexInputStateFactory::ToVkVertexFormat(DataType type, Int size, Bool normalized, Bool isInteger,
+                                                       Bool isBgra) {
+        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
+            // components back into R,G,B,A order for the shader.
+            switch (type) {
+            case DataType::Uint8:
+                return VK_FORMAT_B8G8R8A8_UNORM;
+            case DataType::Uint2101010Rev:
+                return VK_FORMAT_A2R10G10B10_UNORM_PACK32;
+            case DataType::Int2101010Rev:
+                return VK_FORMAT_A2R10G10B10_SNORM_PACK32;
+            default:
+                return VK_FORMAT_UNDEFINED;
+            }
+        }
         switch (type) {
+        case DataType::Uint2101010Rev:
+            // Packed 2_10_10_10 travels the float-normalizing path only; size is always 4. SNORM/UNORM
+            // normalize, SSCALED/USCALED cast the packed field to float.
+            if (isInteger || size != 4) return VK_FORMAT_UNDEFINED;
+            return normalized ? VK_FORMAT_A2B10G10R10_UNORM_PACK32 : VK_FORMAT_A2B10G10R10_USCALED_PACK32;
+        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::Float32:
             switch (size) {
             case 1: return VK_FORMAT_R32_SFLOAT;
@@ -127,6 +214,17 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             case 4: return VK_FORMAT_R32G32B32A32_SFLOAT;
             default: return VK_FORMAT_UNDEFINED;
             }
+        case DataType::Float16:
+            // GL_HALF_FLOAT is a floating-point array type: it is never an integer attribute, and
+            // GL_TRUE for `normalized` is ignored for float types rather than selecting a *NORM format.
+            if (isInteger) return VK_FORMAT_UNDEFINED;
+            switch (size) {
+            case 1: return VK_FORMAT_R16_SFLOAT;
+            case 2: return VK_FORMAT_R16G16_SFLOAT;
+            case 3: return VK_FORMAT_R16G16B16_SFLOAT;
+            case 4: return VK_FORMAT_R16G16B16A16_SFLOAT;
+            default: return VK_FORMAT_UNDEFINED;
+            }
         case DataType::Int32:
             if (!isInteger || normalized) return VK_FORMAT_UNDEFINED;
             switch (size) {
@@ -148,8 +246,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         case DataType::Int16:
             switch (size) {
             case 1:
-                return isInteger ? VK_FORMAT_R16_SINT
-                                 : (normalized ? VK_FORMAT_R16_SNORM : VK_FORMAT_R16_SSCALED);
+                return isInteger ? VK_FORMAT_R16_SINT : (normalized ? VK_FORMAT_R16_SNORM : VK_FORMAT_R16_SSCALED);
             case 2:
                 return isInteger ? VK_FORMAT_R16G16_SINT
                                  : (normalized ? VK_FORMAT_R16G16_SNORM : VK_FORMAT_R16G16_SSCALED);
@@ -164,8 +261,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         case DataType::Uint16:
             switch (size) {
             case 1:
-                return isInteger ? VK_FORMAT_R16_UINT
-                                 : (normalized ? VK_FORMAT_R16_UNORM : VK_FORMAT_R16_USCALED);
+                return isInteger ? VK_FORMAT_R16_UINT : (normalized ? VK_FORMAT_R16_UNORM : VK_FORMAT_R16_USCALED);
             case 2:
                 return isInteger ? VK_FORMAT_R16G16_UINT
                                  : (normalized ? VK_FORMAT_R16G16_UNORM : VK_FORMAT_R16G16_USCALED);
@@ -180,8 +276,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         case DataType::Int8:
             switch (size) {
             case 1:
-                return isInteger ? VK_FORMAT_R8_SINT
-                                 : (normalized ? VK_FORMAT_R8_SNORM : VK_FORMAT_R8_SSCALED);
+                return isInteger ? VK_FORMAT_R8_SINT : (normalized ? VK_FORMAT_R8_SNORM : VK_FORMAT_R8_SSCALED);
             case 2:
                 return isInteger ? VK_FORMAT_R8G8_SINT
                                  : (normalized ? VK_FORMAT_R8G8_SNORM : VK_FORMAT_R8G8_SSCALED);
@@ -196,8 +291,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         case DataType::Uint8:
             switch (size) {
             case 1:
-                return isInteger ? VK_FORMAT_R8_UINT
-                                 : (normalized ? VK_FORMAT_R8_UNORM : VK_FORMAT_R8_USCALED);
+                return isInteger ? VK_FORMAT_R8_UINT : (normalized ? VK_FORMAT_R8_UNORM : VK_FORMAT_R8_USCALED);
             case 2:
                 return isInteger ? VK_FORMAT_R8G8_UINT
                                  : (normalized ? VK_FORMAT_R8G8_UNORM : VK_FORMAT_R8G8_USCALED);
@@ -234,4 +328,57 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return 0;
         }
     }
+
+    SizeT VertexInputStateFactory::GetAttributeByteSize(DataType type, Int size, Bool isBgra) {
+        // The packed 2_10_10_10 types are a single 32-bit word for all 4 components; GL_BGRA is always
+        // 4 components (GL_UNSIGNED_BYTE x4 = 4 bytes, or a packed word = 4 bytes) -- both are 4 bytes.
+        if (type == DataType::Int2101010Rev || type == DataType::Uint2101010Rev || isBgra) {
+            return 4;
+        }
+        const SizeT componentSize = GetComponentSize(type);
+        return componentSize == 0 ? 0 : componentSize * static_cast<SizeT>(size);
+    }
+
+    Bool VertexInputStateFactory::IsScaledIntegerVertexFormat(VkFormat format) {
+        switch (format) {
+        case VK_FORMAT_R8_USCALED:
+        case VK_FORMAT_R8_SSCALED:
+        case VK_FORMAT_R8G8_USCALED:
+        case VK_FORMAT_R8G8_SSCALED:
+        case VK_FORMAT_R8G8B8_USCALED:
+        case VK_FORMAT_R8G8B8_SSCALED:
+        case VK_FORMAT_R8G8B8A8_USCALED:
+        case VK_FORMAT_R8G8B8A8_SSCALED:
+        case VK_FORMAT_R16_USCALED:
+        case VK_FORMAT_R16_SSCALED:
+        case VK_FORMAT_R16G16_USCALED:
+        case VK_FORMAT_R16G16_SSCALED:
+        case VK_FORMAT_R16G16B16_USCALED:
+        case VK_FORMAT_R16G16B16_SSCALED:
+        case VK_FORMAT_R16G16B16A16_USCALED:
+        case VK_FORMAT_R16G16B16A16_SSCALED:
+            return true;
+        default:
+            return false;
+        }
+    }
+
+    VkFormat VertexInputStateFactory::ToFloat32VertexFormat(Int componentCount) {
+        switch (componentCount) {
+        case 1: return VK_FORMAT_R32_SFLOAT;
+        case 2: return VK_FORMAT_R32G32_SFLOAT;
+        case 3: return VK_FORMAT_R32G32B32_SFLOAT;
+        case 4: return VK_FORMAT_R32G32B32A32_SFLOAT;
+        default: return VK_FORMAT_UNDEFINED;
+        }
+    }
+
+    Bool VertexInputStateFactory::SupportsVertexBufferFormat(VkFormat format) const {
+        if (m_physicalDevice == VK_NULL_HANDLE || format == VK_FORMAT_UNDEFINED) {
+            return false;
+        }
+        VkFormatProperties properties{};
+        vkGetPhysicalDeviceFormatProperties(m_physicalDevice, format, &properties);
+        return (properties.bufferFeatures & VK_FORMAT_FEATURE_VERTEX_BUFFER_BIT) != 0;
+    }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VertexInputStateFactory.h

@@ -19,29 +19,56 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     public:
         using HashType = Uint64;
 
+        enum class VertexStreamConversion : Uint8 {
+            None = 0,
+            Repack,
+            ScaledIntegerToFloat32,
+        };
+
         struct BackendVertexInputState {
             HashType hash = 0;
             Vector<VkVertexInputBindingDescription> bindings;
             Vector<VkVertexInputAttributeDescription> attributes;
             Vector<SizeT> bindingBufferKeys;
+            Vector<SizeT> bindingBaseOffsets;
+            Vector<Uint32> bindingAttributeLocations;
+            Vector<Bool> bindingUsesClientMemory;
+            Vector<VertexStreamConversion> bindingConversions;
+            // Locations whose array is ENABLED but whose GL format has no VkFormat mapping. They are
+            // 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;
             VkPipelineVertexInputStateCreateInfo state{
                 VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO
             };
         };
 
-        explicit VertexInputStateFactory(const VulkanRendererConfig& config):
-            m_config(config) {}
+        VertexInputStateFactory(const VulkanRendererConfig& config, VkPhysicalDevice physicalDevice):
+            m_config(config), m_physicalDevice(physicalDevice) {}
         ~VertexInputStateFactory() = default;
         VertexInputStateFactory(const VertexInputStateFactory&) = delete;
 
         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.
+        HashType GetOrComputeHash(const MG_State::GLState::VertexArrayObject& vao) const;
+        const BackendVertexInputState& GetOrCreateVertexInputState(
+            const MG_State::GLState::VertexArrayObject& vao, HashType hash);
         const BackendVertexInputState& GetOrCreateVertexInputState(const MG_State::GLState::VertexArrayObject& vao);
+        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
+        // an unknown/unsupported type.
+        static SizeT GetAttributeByteSize(DataType type, Int size, Bool isBgra);
 
     private:
-        static VkFormat ToVkVertexFormat(DataType type, Int size, Bool normalized, Bool isInteger);
-        static SizeT GetComponentSize(DataType type);
+        static VkFormat ToVkVertexFormat(DataType type, Int size, Bool normalized, Bool isInteger, Bool isBgra = 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;
         static inline XXH64_state_t* m_hashState = XXH64_createState();
     };

MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferManager.cpp

@@ -10,9 +10,73 @@
 
 namespace MobileGL::MG_Backend::DirectVulkan {
     namespace {
-        constexpr Uint32 kResidentBufferGCInterval = 60;
         constexpr VmaAllocationCreateFlags kResidentBufferAllocationFlags =
             VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
+        constexpr SizeT kLiveResourcePruneThreshold = 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.
+        constexpr VkBufferUsageFlags kPersistentBackedUsage =
+            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;
+        // 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 =
+            VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
+
+        using MG_State::GLState::BackendBufferResource;
+        using MG_State::GLState::BufferBackendOps;
+        using MG_State::GLState::BufferObject;
+
+        // The manager owned by the active VulkanRenderer; immediate ops route here.
+        VkBufferManager* g_activeBufferManager = nullptr;
+
+        void Ops_Respecify(BufferObject& bufferObject) {
+            if (g_activeBufferManager) {
+                g_activeBufferManager->OnRespecify(bufferObject);
+            }
+        }
+
+        void Ops_SubData(BufferObject& bufferObject, SizeT offset, SizeT size) {
+            if (g_activeBufferManager) {
+                g_activeBufferManager->OnSubData(bufferObject, offset, size);
+            }
+        }
+
+        void Ops_FlushMappedRange(BufferObject& bufferObject, Range1D range,
+                                  Flags<BufferMappingAccessBit> appAccess) {
+            if (g_activeBufferManager) {
+                g_activeBufferManager->OnFlushMappedRange(bufferObject, range, appAccess);
+            }
+        }
+
+        void* Ops_AcquirePersistentMap(BufferObject& bufferObject) {
+            if (g_activeBufferManager) {
+                return g_activeBufferManager->AcquirePersistentMap(bufferObject);
+            }
+            return nullptr;
+        }
+
+        void Ops_OnDestroy(SharedPtr<BackendBufferResource>&& resource) {
+            if (g_activeBufferManager) {
+                g_activeBufferManager->OnResourceDestroyed(std::move(resource));
+            }
+            // No active manager: the device/allocator is gone or going away and
+            // Shutdown() already destroyed the storage; dropping the handle here
+            // must not touch Vulkan. VkBufferResource's dtor destroys via VMA only
+            // when the allocation is still valid, which Shutdown() cleared.
+        }
+
+        const BufferBackendOps g_vulkanBufferBackendOps = {
+            .Respecify = Ops_Respecify,
+            .SubData = Ops_SubData,
+            .FlushMappedRange = Ops_FlushMappedRange,
+            .OnDestroy = Ops_OnDestroy,
+            .AcquirePersistentMap = Ops_AcquirePersistentMap,
+        };
     } // namespace
 
     Bool VkBufferManager::Initialize(const VkBufferManagerInitInfo& initInfo) {
@@ -22,40 +86,93 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         MOBILEGL_ASSERT(initInfo.frameCount > 0, "VkBufferManager::Initialize requires non-zero frame count");
 
         m_initInfo = initInfo;
-        m_deferredResidentReleases.resize(initInfo.frameCount);
+        m_deferredBufferReleases.resize(initInfo.frameCount);
+        m_deferredResourceReleases.resize(initInfo.frameCount);
         m_currentFrameIndex = 0;
-        return InitializeTransientArenas();
+        m_frameSerial = 1;
+        m_completedSerialFloor = 0;
+        if (!InitializeTransientArenas()) {
+            return false;
+        }
+        g_activeBufferManager = this;
+        MG_State::GLState::SetBufferBackendOps(&g_vulkanBufferBackendOps);
+        return true;
     }
 
     void VkBufferManager::Shutdown() {
+        if (g_activeBufferManager == this) {
+            g_activeBufferManager = nullptr;
+            if (MG_State::GLState::GetBufferBackendOps() == &g_vulkanBufferBackendOps) {
+                MG_State::GLState::SetBufferBackendOps(nullptr);
+            }
+        }
         m_transientUploadArena.Shutdown();
-        DestroyResidentBuffers();
-        DestroyDeferredResidentReleases();
+        DestroyAllDeferredReleases();
+        ReleaseAllLiveResources();
+        m_copyProvider = nullptr;
         m_initInfo = {};
         m_currentFrameIndex = 0;
-        m_residentGcTick = 0;
+        m_frameSerial = 1;
+        m_completedSerialFloor = 0;
     }
 
     Bool VkBufferManager::RecreateTransientArenas(Uint32 frameCount) {
-        MOBILEGL_ASSERT(m_initInfo.allocator != nullptr, "VkBufferManager::RecreateTransientArenas requires initialized manager");
+        MOBILEGL_ASSERT(m_initInfo.allocator != nullptr,
+                        "VkBufferManager::RecreateTransientArenas requires initialized manager");
         MOBILEGL_ASSERT(frameCount > 0, "VkBufferManager::RecreateTransientArenas requires non-zero frame count");
 
+        // Callers guarantee the device is idle around arena recreation.
+        NotifyDeviceIdle();
         m_transientUploadArena.Shutdown();
         m_initInfo.frameCount = frameCount;
-        DestroyDeferredResidentReleases();
-        m_deferredResidentReleases.resize(frameCount);
+        DestroyAllDeferredReleases();
+        m_deferredBufferReleases.resize(frameCount);
+        m_deferredResourceReleases.resize(frameCount);
         m_currentFrameIndex = 0;
         return InitializeTransientArenas();
     }
 
     void VkBufferManager::BeginFrame(Uint32 frameIndex) {
-        MOBILEGL_ASSERT(frameIndex < m_deferredResidentReleases.size(),
+        MOBILEGL_ASSERT(frameIndex < m_deferredBufferReleases.size(),
                         "VkBufferManager::BeginFrame frame index out of range");
         m_currentFrameIndex = frameIndex;
-        CollectDeferredResidentReleases(frameIndex);
+        ++m_frameSerial;
+        CollectDeferredReleases(frameIndex);
         m_transientUploadArena.BeginFrame(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
+        // remains busy.
+        if (m_frameSerial > 0) {
+            m_completedSerialFloor = m_frameSerial - 1;
+        }
+    }
+
+    void VkBufferManager::NotifyFrameSerialComplete(Uint64 serial) {
+        // The current serial's work is still being recorded; a completion
+        // report for it (or beyond) can only come from a stale caller.
+        if (serial >= m_frameSerial) {
+            return;
+        }
+        m_completedSerialFloor = std::max(m_completedSerialFloor, serial);
+    }
+
+    void VkBufferManager::SetCopyCommandProvider(IBufferCopyCommandProvider* provider) {
+        m_copyProvider = provider;
+    }
+
+    Uint64 VkBufferManager::GetCompletedSerial() const {
+        const Uint64 frameCount = m_initInfo.frameCount > 0 ? m_initInfo.frameCount : 1;
+        const Uint64 completed = m_frameSerial > frameCount ? m_frameSerial - frameCount : 0;
+        return std::max(completed, m_completedSerialFloor);
+    }
+
+    Bool VkBufferManager::IsResourceBusy(const VkBufferResource& resource) const {
+        return resource.lastUseSerial > GetCompletedSerial();
+    }
+
     Bool VkBufferManager::UploadTransient(BufferKind kind, Uint32 frameIndex, const void* data,
                                           VkDeviceSize size, VkDeviceSize alignment, BufferSlice& outSlice) {
         (void)kind;
@@ -67,7 +184,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             .allocator = m_initInfo.allocator,
             .frameCount = m_initInfo.frameCount,
             .usage = VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT |
-                     VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT,
+                     VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT |
+                     VK_BUFFER_USAGE_TRANSFER_SRC_BIT,
             .memoryUsage = m_initInfo.transientMemoryUsage,
             .allocationFlags = m_initInfo.transientAllocationFlags,
             .minBufferSize = m_initInfo.minUploadBytes,
@@ -75,115 +193,394 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         });
     }
 
-    Bool VkBufferManager::SyncResidentBuffer(BufferKind kind,
-                                             const SharedPtr<MG_State::GLState::BufferObject>& bufferObject,
-                                             BufferSlice& outSlice) {
-        const VkBufferUsageFlags requiredUsage = GetVkBufferUsage(kind);
-        MOBILEGL_ASSERT(requiredUsage != 0,
-                        "VkBufferManager::SyncResidentBuffer only supports resident vertex/index buffers");
-        MOBILEGL_ASSERT(bufferObject != nullptr, "VkBufferManager::SyncResidentBuffer requires valid buffer object");
-        CollectResidentGarbageIfNeeded();
+    VkBufferResource* VkBufferManager::ResourceOf(MG_State::GLState::BufferObject& bufferObject) {
+        return static_cast<VkBufferResource*>(bufferObject.GetBackendResource().get());
+    }
 
-        const auto* bufferData = bufferObject->GetDataReadOnly().get();
-        MOBILEGL_ASSERT(bufferData != nullptr, "VkBufferManager::SyncResidentBuffer requires frontend buffer data");
+    VkBufferResource* VkBufferManager::GetOrCreateResource(
+        const SharedPtr<MG_State::GLState::BufferObject>& bufferObject) {
+        // Return by raw pointer: the resource is owned for its whole lifetime by the BufferObject's
+        // backend-resource SharedPtr (already set, or set below), so callers that only dereference
+        // it avoid a static_pointer_cast + SharedPtr refcount inc/dec on every per-draw buffer bind.
+        const auto& existing = bufferObject->GetBackendResource();
+        if (existing) {
+            return static_cast<VkBufferResource*>(existing.get());
+        }
+        auto resource = MakeShared<VkBufferResource>();
+        VkBufferResource* raw = resource.get();
+        bufferObject->SetBackendResource(resource);
+        TrackLiveResource(resource);
+        return raw;
+    }
 
-        const VkDeviceSize bufferSize = static_cast<VkDeviceSize>(bufferObject->GetSize());
-        if (bufferSize == 0) {
-            MGLOG_E("VkBufferManager::SyncResidentBuffer failed: buffer size is zero");
+    void VkBufferManager::TrackLiveResource(const SharedPtr<VkBufferResource>& resource) {
+        if (m_liveResources.size() >= kLiveResourcePruneThreshold) {
+            std::erase_if(m_liveResources, [](const WeakPtr<VkBufferResource>& weak) { return weak.expired(); });
+        }
+        m_liveResources.push_back(resource);
+    }
+
+    void VkBufferManager::ReleaseAllLiveResources() {
+        for (auto& weak : m_liveResources) {
+            if (auto resource = weak.lock()) {
+                resource->buffer.Destroy();
+                resource->storageSize = 0;
+                resource->usageFlags = 0;
+                resource->lastUseSerial = 0;
+                resource->pendingFullUpload = true;
+                resource->transientSlice = {};
+                resource->transientFrameSerial = 0;
+            }
+        }
+        m_liveResources.clear();
+    }
+
+    Bool VkBufferManager::CreateResidentStorage(VkBufferResource& resource, VkDeviceSize size,
+                                                VkBufferUsageFlags usage, VkMemoryPropertyFlags requiredFlags) {
+        // Staged range copies write resident storage with vkCmdCopyBuffer.
+        usage |= VK_BUFFER_USAGE_TRANSFER_DST_BIT;
+        const Bool created = resource.buffer.Create({
+            .allocator = m_initInfo.allocator,
+            .size = size,
+            .usage = usage,
+            .memoryUsage = VMA_MEMORY_USAGE_AUTO,
+            .allocationFlags = kResidentBufferAllocationFlags,
+            .requiredFlags = requiredFlags,
+        });
+        if (!created || resource.buffer.Map() == nullptr) {
+            MGLOG_E("VkBufferManager::CreateResidentStorage failed (size=%llu)",
+                    static_cast<unsigned long long>(size));
+            resource.buffer.Destroy();
+            resource.storageSize = 0;
+            resource.usageFlags = 0;
             return false;
         }
-
-        auto& entry = m_residentBuffers[bufferObject.get()];
-        entry.aliveRef = bufferObject;
-
-        const auto changeBits = bufferObject->GetChangeBits();
-        const Bool needsRecreate = !entry.buffer.IsValid() || entry.size != bufferSize ||
-                                   ((entry.usage & requiredUsage) != requiredUsage) ||
-                                   (changeBits & BufferChangeBits::PreferReallocationBit);
-        if (needsRecreate) {
-            const VkBufferUsageFlags recreatedUsage = entry.usage | requiredUsage;
-            DeferResidentRelease(std::move(entry.buffer));
-            const Bool created = entry.buffer.Create({
-                .allocator = m_initInfo.allocator,
-                .size = bufferSize,
-                .usage = recreatedUsage,
-                .memoryUsage = VMA_MEMORY_USAGE_AUTO,
-                .allocationFlags = kResidentBufferAllocationFlags,
-            });
-            if (!created || entry.buffer.Map() == nullptr) {
-                MGLOG_E("VkBufferManager::SyncResidentBuffer failed: unable to create resident buffer");
-                entry.buffer.Destroy();
-                entry.size = 0;
-                entry.usage = 0;
-                return false;
-            }
-            if (!entry.buffer.Upload(bufferData->data(), bufferSize, 0)) {
-                MGLOG_E("VkBufferManager::SyncResidentBuffer failed: initial upload failed");
-                entry.buffer.Destroy();
-                entry.size = 0;
-                entry.usage = 0;
-                return false;
-            }
-            entry.size = bufferSize;
-            entry.usage = recreatedUsage;
-            bufferObject->ClearDirty();
-            outSlice = entry.buffer.GetSlice(0, bufferSize);
-            return true;
-        }
-
-        if (changeBits & BufferChangeBits::DirtyBit) {
-            const auto& dirtyRanges = bufferObject->GetDirtyRanges();
-            for (const auto& range : dirtyRanges) {
-                const VkDeviceSize rangeOffset = static_cast<VkDeviceSize>(range.start);
-                const VkDeviceSize rangeSize = static_cast<VkDeviceSize>(range.end - range.start);
-                if (rangeSize == 0) {
-                    continue;
-                }
-                if (!entry.buffer.Upload(bufferData->data() + range.start, rangeSize, rangeOffset)) {
-                    MGLOG_E("VkBufferManager::SyncResidentBuffer failed: dirty range upload failed");
-                    return false;
-                }
-            }
-            bufferObject->ClearDirty();
-        }
-
-        outSlice = entry.buffer.GetSlice(0, bufferSize);
+        resource.storageSize = size;
+        resource.usageFlags = usage;
         return true;
     }
 
-    void VkBufferManager::DowngradeResidentBufferToTransient(const SharedPtr<MG_State::GLState::BufferObject>& bufferObject) {
-        if (bufferObject == nullptr) {
-            return;
+    Bool VkBufferManager::SwapStorageAndUploadAll(VkBufferResource& resource,
+                                                  MG_State::GLState::BufferObject& bufferObject) {
+        const VkDeviceSize size = static_cast<VkDeviceSize>(bufferObject.GetSize());
+        const VkBufferUsageFlags usage = resource.usageFlags;
+        DeferRelease(std::move(resource.buffer));
+        if (!CreateResidentStorage(resource, size, usage)) {
+            resource.pendingFullUpload = true;
+            return false;
         }
-
-        auto it = m_residentBuffers.find(bufferObject.get());
-        if (it == m_residentBuffers.end()) {
-            return;
+        if (!resource.buffer.Upload(bufferObject.MappedData(), size, 0)) {
+            MGLOG_E("VkBufferManager::SwapStorageAndUploadAll: upload failed");
+            resource.pendingFullUpload = true;
+            return false;
         }
-
-        DeferResidentRelease(std::move(it->second.buffer));
-        m_residentBuffers.erase(it);
+        resource.pendingFullUpload = false;
+        return true;
     }
 
-    void VkBufferManager::DeferResidentRelease(VkBufferObject&& buffer) {
+    Bool VkBufferManager::StagedRangeCopy(VkBufferResource& resource, MG_State::GLState::BufferObject& bufferObject,
+                                          SizeT offset, SizeT size) {
+        if (!m_copyProvider) {
+            return false;
+        }
+        BufferSlice staging{};
+        if (!m_transientUploadArena.Upload(m_currentFrameIndex, bufferObject.MappedData() + offset,
+                                           static_cast<VkDeviceSize>(size), 16, staging)) {
+            return false;
+        }
+        VkCommandBuffer commandBuffer = m_copyProvider->AcquireBufferCopyCommandBuffer();
+        if (commandBuffer == VK_NULL_HANDLE) {
+            return false;
+        }
+
+        // Order the copy after every prior read/write of this buffer, both from
+        // in-flight frames (submission order) and from commands already recorded
+        // in this frame's command buffer.
+        VkMemoryBarrier beforeBarrier{VK_STRUCTURE_TYPE_MEMORY_BARRIER};
+        beforeBarrier.srcAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT;
+        beforeBarrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
+        vkCmdPipelineBarrier(commandBuffer, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 1,
+                             &beforeBarrier, 0, nullptr, 0, nullptr);
+
+        VkBufferCopy region{};
+        region.srcOffset = staging.offset;
+        region.dstOffset = static_cast<VkDeviceSize>(offset);
+        region.size = static_cast<VkDeviceSize>(size);
+        vkCmdCopyBuffer(commandBuffer, staging.buffer, resource.buffer.GetHandle(), 1, &region);
+
+        VkMemoryBarrier afterBarrier{VK_STRUCTURE_TYPE_MEMORY_BARRIER};
+        afterBarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
+        afterBarrier.dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT;
+        vkCmdPipelineBarrier(commandBuffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, 1,
+                             &afterBarrier, 0, nullptr, 0, nullptr);
+
+        resource.lastUseSerial = m_frameSerial;
+        return true;
+    }
+
+    void VkBufferManager::OnRespecify(MG_State::GLState::BufferObject& bufferObject) {
+        auto* resource = ResourceOf(bufferObject);
+        if (!resource) {
+            return; // lazy: AcquireResidentSlice performs a full upload on creation
+        }
+        // Any cached streaming slice refers to the previous contents.
+        resource->transientFrameSerial = 0;
+        if (!resource->buffer.IsValid()) {
+            return; // streaming-only resource: shadow + serial are enough
+        }
+
+        const VkDeviceSize size = static_cast<VkDeviceSize>(bufferObject.GetSize());
+        if (size == 0) {
+            DeferRelease(std::move(resource->buffer));
+            resource->storageSize = 0;
+            resource->pendingFullUpload = false;
+            return;
+        }
+
+        if (size != resource->storageSize || IsResourceBusy(*resource)) {
+            // Conditional orphan: only swap the storage when the old one is
+            // still referenced by the GPU (or no longer fits).
+            SwapStorageAndUploadAll(*resource, bufferObject);
+            return;
+        }
+
+        if (!resource->buffer.Upload(bufferObject.MappedData(), size, 0)) {
+            MGLOG_E("VkBufferManager::OnRespecify: in-place upload failed");
+            resource->pendingFullUpload = true;
+        }
+    }
+
+    void VkBufferManager::OnSubData(MG_State::GLState::BufferObject& bufferObject, SizeT offset, SizeT size) {
+        auto* resource = ResourceOf(bufferObject);
+        if (!resource) {
+            return;
+        }
+        resource->transientFrameSerial = 0;
+        if (!resource->buffer.IsValid() || resource->pendingFullUpload) {
+            return;
+        }
+        if (static_cast<VkDeviceSize>(bufferObject.GetSize()) != resource->storageSize) {
+            resource->pendingFullUpload = true;
+            return;
+        }
+
+        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");
+                resource->pendingFullUpload = true;
+            }
+            return;
+        }
+
+        // Busy partial write: stage + GPU copy preserves GL ordering within the
+        // frame and leaves bytes outside the range (possibly GPU-written, e.g.
+        // SSBO) intact. Fall back to a storage swap if staging is unavailable.
+        if (!StagedRangeCopy(*resource, bufferObject, offset, size)) {
+            SwapStorageAndUploadAll(*resource, bufferObject);
+        }
+    }
+
+    void VkBufferManager::OnFlushMappedRange(MG_State::GLState::BufferObject& bufferObject, Range1D range,
+                                             Flags<BufferMappingAccessBit> appAccess) {
+        auto* resource = ResourceOf(bufferObject);
+        if (!resource) {
+            return;
+        }
+        resource->transientFrameSerial = 0;
+        if (!resource->buffer.IsValid() || resource->pendingFullUpload) {
+            return;
+        }
+        if (static_cast<VkDeviceSize>(bufferObject.GetSize()) != resource->storageSize) {
+            resource->pendingFullUpload = true;
+            return;
+        }
+
+        const SizeT offset = range.start;
+        const SizeT size = range.end - range.start;
+        // GL_MAP_UNSYNCHRONIZED_BIT: the app guarantees it does not overwrite
+        // data the GPU is still reading; honour it with a direct host write.
+        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");
+                resource->pendingFullUpload = true;
+            }
+            return;
+        }
+
+        if (!StagedRangeCopy(*resource, bufferObject, offset, size)) {
+            SwapStorageAndUploadAll(*resource, bufferObject);
+        }
+    }
+
+    void VkBufferManager::OnResourceDestroyed(SharedPtr<MG_State::GLState::BackendBufferResource>&& resource) {
+        if (!resource) {
+            return;
+        }
+        auto vkResource = std::static_pointer_cast<VkBufferResource>(std::move(resource));
+        if (!vkResource->buffer.IsValid()) {
+            return;
+        }
+        if (m_deferredResourceReleases.empty()) {
+            vkResource->buffer.Destroy();
+            return;
+        }
+        MOBILEGL_ASSERT(m_currentFrameIndex < m_deferredResourceReleases.size(),
+                        "VkBufferManager::OnResourceDestroyed current frame index out of range");
+        // Keep the whole resource alive until this frame slot's fence has been
+        // waited, then the storage is destroyed with it.
+        m_deferredResourceReleases[m_currentFrameIndex].push_back(std::move(vkResource));
+    }
+
+    void* VkBufferManager::AcquirePersistentMap(MG_State::GLState::BufferObject& bufferObject) {
+        const VkDeviceSize size = static_cast<VkDeviceSize>(bufferObject.GetSize());
+        if (size == 0) {
+            return nullptr;
+        }
+
+        auto resource = std::static_pointer_cast<VkBufferResource>(bufferObject.GetBackendResource());
+        if (!resource) {
+            resource = MakeShared<VkBufferResource>();
+            bufferObject.SetBackendResource(resource);
+            TrackLiveResource(resource);
+        }
+
+        // Idempotent: an already-backed buffer returns the same mapped base.
+        if (resource->persistentMapped && resource->buffer.IsValid() && resource->storageSize == size) {
+            return resource->buffer.GetMappedData();
+        }
+
+        // One-time creation of HOST_VISIBLE + HOST_COHERENT, persistently mapped storage
+        // carrying every usage (never recreated, so the app's pointer never dangles). Seed
+        // 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)) {
+            resource->persistentMapped = false;
+            resource->storageSize = 0;
+            resource->usageFlags = 0;
+            return nullptr;
+        }
+        const Uint8* seed = bufferObject.MappedData();
+        if (seed != nullptr) {
+            resource->buffer.Upload(seed, size, 0);
+        }
+        resource->persistentMapped = true;
+        resource->pendingFullUpload = false;
+        resource->storageSize = size;
+        resource->lastUseSerial = 0;
+        return resource->buffer.GetMappedData();
+    }
+
+    Bool VkBufferManager::AcquireResidentSlice(BufferKind kind,
+                                               const SharedPtr<MG_State::GLState::BufferObject>& bufferObject,
+                                               BufferSlice& outSlice) {
+        const VkBufferUsageFlags requiredUsage = GetVkBufferUsage(kind);
+        MOBILEGL_ASSERT(requiredUsage != 0, "VkBufferManager::AcquireResidentSlice unsupported buffer kind");
+        MOBILEGL_ASSERT(bufferObject != nullptr, "VkBufferManager::AcquireResidentSlice requires valid buffer object");
+
+        auto resource = GetOrCreateResource(bufferObject);
+        bufferObject->SyncPersistentMappedRange();
+
+        const VkDeviceSize size = static_cast<VkDeviceSize>(bufferObject->GetSize());
+        if (size == 0) {
+            MGLOG_E("VkBufferManager::AcquireResidentSlice failed: buffer size is zero");
+            return false;
+        }
+
+        // Zero-copy persistent buffers already hold the app's live coherent writes in
+        // host-visible storage carrying every usage; bind directly, no re-upload/staging.
+        if (resource->persistentMapped && resource->buffer.IsValid() && resource->storageSize == size) {
+            resource->lastUseSerial = m_frameSerial;
+            outSlice = resource->buffer.GetSlice(0, size);
+            return outSlice.IsValid();
+        }
+
+        const Bool needsRecreate = !resource->buffer.IsValid() || resource->storageSize != size ||
+                                   ((resource->usageFlags & requiredUsage) != requiredUsage) ||
+                                   resource->pendingFullUpload;
+        if (needsRecreate) {
+            const VkBufferUsageFlags usage = resource->usageFlags | requiredUsage;
+            DeferRelease(std::move(resource->buffer));
+            if (!CreateResidentStorage(*resource, size, usage)) {
+                return false;
+            }
+            if (!resource->buffer.Upload(bufferObject->MappedData(), size, 0)) {
+                MGLOG_E("VkBufferManager::AcquireResidentSlice failed: initial upload failed");
+                resource->buffer.Destroy();
+                resource->storageSize = 0;
+                resource->usageFlags = 0;
+                return false;
+            }
+            resource->pendingFullUpload = false;
+        }
+
+        resource->lastUseSerial = m_frameSerial;
+        outSlice = resource->buffer.GetSlice(0, size);
+        return true;
+    }
+
+    Bool VkBufferManager::AcquireStreamedSlice(BufferKind kind,
+                                               const SharedPtr<MG_State::GLState::BufferObject>& bufferObject,
+                                               BufferSlice& outSlice) {
+        (void)kind;
+        MOBILEGL_ASSERT(bufferObject != nullptr, "VkBufferManager::AcquireStreamedSlice requires valid buffer object");
+
+        auto resource = GetOrCreateResource(bufferObject);
+        bufferObject->SyncPersistentMappedRange();
+
+        const VkDeviceSize size = static_cast<VkDeviceSize>(bufferObject->GetSize());
+        if (size == 0) {
+            MGLOG_E("VkBufferManager::AcquireStreamedSlice failed: buffer size is zero");
+            return false;
+        }
+
+        const Uint64 changeSerial = bufferObject->GetChangeSerial();
+        if (resource->transientFrameSerial == m_frameSerial && resource->transientChangeSerial == changeSerial &&
+            resource->transientSize == size && resource->transientSlice.IsValid()) {
+            outSlice = resource->transientSlice;
+            return true;
+        }
+
+        if (!m_transientUploadArena.Upload(m_currentFrameIndex, bufferObject->MappedData(), size, 16,
+                                           outSlice)) {
+            return false;
+        }
+        resource->transientSlice = outSlice;
+        resource->transientFrameSerial = m_frameSerial;
+        resource->transientChangeSerial = changeSerial;
+        resource->transientSize = size;
+
+        // Streaming path is authoritative now; release resident storage so we do
+        // not keep a second, stale copy alive (downgrade).
+        if (resource->buffer.IsValid()) {
+            DeferRelease(std::move(resource->buffer));
+            resource->storageSize = 0;
+        }
+        return true;
+    }
+
+    void VkBufferManager::DeferRelease(VkBufferObject&& buffer) {
         if (!buffer.IsValid()) {
             return;
         }
 
-        if (m_deferredResidentReleases.empty()) {
+        if (m_deferredBufferReleases.empty()) {
             buffer.Destroy();
             return;
         }
 
-        MOBILEGL_ASSERT(m_currentFrameIndex < m_deferredResidentReleases.size(),
-                        "VkBufferManager::DeferResidentRelease current frame index out of range");
-        m_deferredResidentReleases[m_currentFrameIndex].push_back(std::move(buffer));
+        MOBILEGL_ASSERT(m_currentFrameIndex < m_deferredBufferReleases.size(),
+                        "VkBufferManager::DeferRelease current frame index out of range");
+        m_deferredBufferReleases[m_currentFrameIndex].push_back(std::move(buffer));
     }
 
-    void VkBufferManager::CollectDeferredResidentReleases(Uint32 frameIndex) {
-        MOBILEGL_ASSERT(frameIndex < m_deferredResidentReleases.size(),
-                        "VkBufferManager::CollectDeferredResidentReleases frame index out of range");
-        m_deferredResidentReleases[frameIndex].clear();
+    void VkBufferManager::CollectDeferredReleases(Uint32 frameIndex) {
+        MOBILEGL_ASSERT(frameIndex < m_deferredBufferReleases.size(),
+                        "VkBufferManager::CollectDeferredReleases frame index out of range");
+        m_deferredBufferReleases[frameIndex].clear();
+        m_deferredResourceReleases[frameIndex].clear();
     }
 
     VkBufferUsageFlags VkBufferManager::GetVkBufferUsage(BufferKind kind) {
@@ -196,51 +593,32 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             // never need to recreate a buffer after it has already been bound.
             return VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT;
         case BufferKind::Uniform:
+            return VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT;
+        case BufferKind::TextureBuffer:
+            return VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT;
+        case BufferKind::ShaderStorage:
+            return VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT;
+        case BufferKind::Indirect:
+            return VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
         default:
             return 0;
         }
     }
 
-    void VkBufferManager::CollectResidentGarbageIfNeeded() {
-        ++m_residentGcTick;
-        if (m_residentGcTick < kResidentBufferGCInterval) {
-            return;
-        }
-        CollectResidentGarbageNow();
-        m_residentGcTick = 0;
-    }
-
-    void VkBufferManager::CollectResidentGarbageNow() {
-        Vector<MG_State::GLState::BufferObject*> staleBuffers;
-        staleBuffers.reserve(m_residentBuffers.size());
-
-        for (const auto& [rawBuffer, entry] : m_residentBuffers) {
-            if (entry.aliveRef.expired()) {
-                staleBuffers.push_back(rawBuffer);
+    void VkBufferManager::DestroyAllDeferredReleases() {
+        for (auto& releases : m_deferredBufferReleases) {
+            for (auto& buffer : releases) {
+                buffer.Destroy();
             }
+            releases.clear();
         }
-
-        for (const auto* rawBuffer : staleBuffers) {
-            auto it = m_residentBuffers.find(const_cast<MG_State::GLState::BufferObject*>(rawBuffer));
-            if (it == m_residentBuffers.end()) {
-                continue;
+        m_deferredBufferReleases.clear();
+        for (auto& releases : m_deferredResourceReleases) {
+            for (auto& resource : releases) {
+                resource->buffer.Destroy();
             }
-            DeferResidentRelease(std::move(it->second.buffer));
-            m_residentBuffers.erase(it);
+            releases.clear();
         }
-    }
-
-    void VkBufferManager::DestroyDeferredResidentReleases() {
-        for (auto& deferredReleases : m_deferredResidentReleases) {
-            deferredReleases.clear();
-        }
-        m_deferredResidentReleases.clear();
-    }
-
-    void VkBufferManager::DestroyResidentBuffers() {
-        for (auto& [_, entry] : m_residentBuffers) {
-            entry.buffer.Destroy();
-        }
-        m_residentBuffers.clear();
+        m_deferredResourceReleases.clear();
     }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferManager.h

@@ -19,6 +19,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Vertex,
         Index,
         Uniform,
+        TextureBuffer,
+        ShaderStorage,
+        Indirect,
     };
 
     struct VkBufferManagerInitInfo {
@@ -30,6 +33,42 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Bool transientPersistentMapping = false;
     };
 
+    // The DirectVulkan storage behind one frontend buffer (pipe_resource analogue).
+    // Owned (refcounted) by the frontend BufferObject; the manager holds only weak
+    // references (for shutdown) plus strong references on deferred-release lists.
+    class VkBufferResource : public MG_State::GLState::BackendBufferResource {
+    public:
+        ~VkBufferResource() override = default;
+
+        // Resident storage (may be invalid for streaming-only buffers).
+        VkBufferObject buffer;
+        VkDeviceSize storageSize = 0;
+        VkBufferUsageFlags usageFlags = 0;
+        // Frame serial of the last GPU reference; drives busy tracking.
+        Uint64 lastUseSerial = 0;
+        // Set when an immediate op could not be applied; forces a full re-upload
+        // on the next AcquireResidentSlice.
+        Bool pendingFullUpload = false;
+        // Backs a zero-copy coherent persistent map (PipeResource GPU residency): the
+        // buffer is HOST_VISIBLE+COHERENT, persistently mapped, carries every usage and is
+        // never orphaned or recreated. Draw-time acquire binds it directly, no re-upload.
+        Bool persistentMapped = false;
+
+        // Cached transient (streaming) slice for the current frame.
+        BufferSlice transientSlice{};
+        Uint64 transientFrameSerial = 0;
+        Uint64 transientChangeSerial = 0;
+        VkDeviceSize transientSize = 0;
+    };
+
+    // Supplies a command buffer that is recording and outside any render pass,
+    // for staged buffer-range copies. Implemented by VulkanRenderer.
+    class IBufferCopyCommandProvider {
+    public:
+        virtual ~IBufferCopyCommandProvider() = default;
+        virtual VkCommandBuffer AcquireBufferCopyCommandBuffer() = 0;
+    };
+
     class VkBufferManager {
     public:
         Bool Initialize(const VkBufferManagerInitInfo& initInfo);
@@ -38,35 +77,77 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         // Recreate all per-frame transient arenas
         Bool RecreateTransientArenas(Uint32 frameCount);
         void BeginFrame(Uint32 frameIndex);
+        // All previously submitted GPU work has completed (vkDeviceWaitIdle).
+        void NotifyDeviceIdle();
+        // A frame slot's submission fence has been waited: every serial up to
+        // and including `serial` is complete. Raises the completed floor so
+        // GetCompletedSerial reflects real fence progress instead of only the
+        // frameSerial-minus-frameCount inference.
+        void NotifyFrameSerialComplete(Uint64 serial);
+        void SetCopyCommandProvider(IBufferCopyCommandProvider* provider);
 
         Bool UploadTransient(BufferKind kind, Uint32 frameIndex, const void* data, VkDeviceSize size,
                              VkDeviceSize alignment, BufferSlice& outSlice);
-        Bool SyncResidentBuffer(BufferKind kind, const SharedPtr<MG_State::GLState::BufferObject>& bufferObject,
-                                BufferSlice& outSlice);
-        void DowngradeResidentBufferToTransient(const SharedPtr<MG_State::GLState::BufferObject>& bufferObject);
+
+        // 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,
+                                  BufferSlice& outSlice);
+        // Draw-time acquire for streamed buffers: uploads the whole shadow into
+        // the per-frame arena (cached by change serial), releasing any resident
+        // storage the buffer may still own.
+        Bool AcquireStreamedSlice(BufferKind kind, const SharedPtr<MG_State::GLState::BufferObject>& bufferObject,
+                                  BufferSlice& outSlice);
+
+        // Zero-copy persistent map (PipeResource GPU residency): create (once) a
+        // HOST_VISIBLE+COHERENT, persistently mapped resident buffer carrying every usage,
+        // seed it from the shadow, and return its mapped base for the app to write into
+        // directly. Idempotent. Returns nullptr on failure (frontend keeps its shadow).
+        void* AcquirePersistentMap(MG_State::GLState::BufferObject& bufferObject);
+
+        // Immediate ops, dispatched from the frontend BufferBackendOps table.
+        void OnRespecify(MG_State::GLState::BufferObject& bufferObject);
+        void OnSubData(MG_State::GLState::BufferObject& bufferObject, SizeT offset, SizeT size);
+        void OnFlushMappedRange(MG_State::GLState::BufferObject& bufferObject, Range1D range,
+                                Flags<BufferMappingAccessBit> appAccess);
+        void OnResourceDestroyed(SharedPtr<MG_State::GLState::BackendBufferResource>&& resource);
+
+        Uint64 GetFrameSerial() const { return m_frameSerial; }
+        // 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.
+        Uint64 GetCompletedSerial() const;
+        // Busy = potentially referenced by GPU work that has not been fenced yet
+        // (including commands recorded for the current, unsubmitted frame).
+        Bool IsResourceBusy(const VkBufferResource& resource) const;
 
     private:
-        struct ResidentBufferEntry {
-            WeakPtr<MG_State::GLState::BufferObject> aliveRef;
-            VkBufferObject buffer;
-            VkDeviceSize size = 0;
-            VkBufferUsageFlags usage = 0;
-        };
-
         Bool InitializeTransientArenas();
         static VkBufferUsageFlags GetVkBufferUsage(BufferKind kind);
-        void DeferResidentRelease(VkBufferObject&& buffer);
-        void CollectDeferredResidentReleases(Uint32 frameIndex);
-        void CollectResidentGarbageIfNeeded();
-        void CollectResidentGarbageNow();
-        void DestroyDeferredResidentReleases();
-        void DestroyResidentBuffers();
+        VkBufferResource* GetOrCreateResource(const SharedPtr<MG_State::GLState::BufferObject>& bufferObject);
+        static VkBufferResource* ResourceOf(MG_State::GLState::BufferObject& bufferObject);
+        Bool CreateResidentStorage(VkBufferResource& resource, VkDeviceSize size, VkBufferUsageFlags usage,
+                                   VkMemoryPropertyFlags requiredFlags = 0);
+        // Swap storage (conditional orphan) and refill it from the shadow copy.
+        Bool SwapStorageAndUploadAll(VkBufferResource& resource, MG_State::GLState::BufferObject& bufferObject);
+        // Record a staging-slice copy into the resident storage, ordered against
+        // in-flight and already-recorded GPU work.
+        Bool StagedRangeCopy(VkBufferResource& resource, MG_State::GLState::BufferObject& bufferObject,
+                             SizeT offset, SizeT size);
+        void DeferRelease(VkBufferObject&& buffer);
+        void CollectDeferredReleases(Uint32 frameIndex);
+        void DestroyAllDeferredReleases();
+        void TrackLiveResource(const SharedPtr<VkBufferResource>& resource);
+        void ReleaseAllLiveResources();
 
         VkBufferManagerInitInfo m_initInfo{};
         BufferArena m_transientUploadArena;
-        UnorderedMap<MG_State::GLState::BufferObject*, ResidentBufferEntry> m_residentBuffers;
-        Vector<Vector<VkBufferObject>> m_deferredResidentReleases;
+        IBufferCopyCommandProvider* m_copyProvider = nullptr;
+        Vector<Vector<VkBufferObject>> m_deferredBufferReleases;
+        Vector<Vector<SharedPtr<VkBufferResource>>> m_deferredResourceReleases;
+        Vector<WeakPtr<VkBufferResource>> m_liveResources;
         Uint32 m_currentFrameIndex = 0;
-        Uint32 m_residentGcTick = 0;
+        Uint64 m_frameSerial = 1;
+        Uint64 m_completedSerialFloor = 0;
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferObject.cpp

@@ -49,11 +49,13 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     }
 
     Bool VkBufferObject::Create(const VkBufferObjectDesc& desc) {
-        return Create(desc.allocator, desc.size, desc.usage, desc.memoryUsage, desc.allocationFlags);
+        return Create(desc.allocator, desc.size, desc.usage, desc.memoryUsage, desc.allocationFlags,
+                      desc.requiredFlags);
     }
 
     Bool VkBufferObject::Create(VmaAllocator allocator, VkDeviceSize size, VkBufferUsageFlags usage,
-                                VmaMemoryUsage memoryUsage, VmaAllocationCreateFlags allocationFlags) {
+                                VmaMemoryUsage memoryUsage, VmaAllocationCreateFlags allocationFlags,
+                                VkMemoryPropertyFlags requiredFlags) {
         MOBILEGL_ASSERT(allocator != nullptr, "VkBufferObject::Create requires valid VMA allocator");
         MOBILEGL_ASSERT(size > 0, "VkBufferObject::Create requires non-zero size");
 
@@ -69,6 +71,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VmaAllocationCreateInfo allocationInfo{};
         allocationInfo.usage = memoryUsage;
         allocationInfo.flags = allocationFlags;
+        allocationInfo.requiredFlags = requiredFlags;
 
         const VkResult result =
             vmaCreateBuffer(m_allocator, &bufferInfo, &allocationInfo, &m_buffer, &m_allocation, nullptr);
@@ -140,12 +143,39 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
 
         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);
+            if (!wasMapped) {
+                Unmap();
+            }
+            return false;
+        }
         if (!wasMapped) {
             Unmap();
         }
         return true;
     }
 
+    Bool VkBufferObject::Invalidate(VkDeviceSize size, VkDeviceSize offset) {
+        MOBILEGL_ASSERT(IsValid(), "VkBufferObject::Invalidate called on invalid buffer");
+        MOBILEGL_ASSERT(IsMapped(), "VkBufferObject::Invalidate requires mapped memory");
+        MOBILEGL_ASSERT(offset <= m_size, "VkBufferObject::Invalidate offset out of range");
+
+        const VkDeviceSize resolvedSize = size == VK_WHOLE_SIZE ? m_size - offset : size;
+        MOBILEGL_ASSERT(offset + resolvedSize <= m_size, "VkBufferObject::Invalidate range out of bounds");
+        if (resolvedSize == 0) {
+            return true;
+        }
+
+        const VkResult result = vmaInvalidateAllocation(m_allocator, m_allocation, offset, resolvedSize);
+        if (result != VK_SUCCESS) {
+            MGLOG_E("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/MG_Backend/DirectVulkan/Renderer/VkBufferObject.h

@@ -20,6 +20,9 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkBufferUsageFlags usage = 0;
         VmaMemoryUsage memoryUsage = VMA_MEMORY_USAGE_AUTO;
         VmaAllocationCreateFlags allocationFlags = 0;
+        // Memory property bits the allocation MUST satisfy (e.g. HOST_VISIBLE|HOST_COHERENT
+        // for a persistently-mapped buffer the app writes into without explicit flushes).
+        VkMemoryPropertyFlags requiredFlags = 0;
     };
 
     class VkBufferObject {
@@ -34,12 +37,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         Bool Create(const VkBufferObjectDesc& desc);
         Bool Create(VmaAllocator allocator, VkDeviceSize size, VkBufferUsageFlags usage,
-                    VmaMemoryUsage memoryUsage, VmaAllocationCreateFlags allocationFlags = 0);
+                    VmaMemoryUsage memoryUsage, VmaAllocationCreateFlags allocationFlags = 0,
+                    VkMemoryPropertyFlags requiredFlags = 0);
         void Destroy();
 
         void* Map();
         void Unmap();
         Bool Upload(const void* data, VkDeviceSize size, VkDeviceSize offset = 0);
+        Bool Invalidate(VkDeviceSize size = VK_WHOLE_SIZE, VkDeviceSize offset = 0);
 
         VkBuffer GetHandle() const { return m_buffer; }
         VkDeviceSize GetSize() const { return m_size; }

MobileGL/MG_Backend/DirectVulkan/Renderer/VkClearManager.cpp

@@ -12,12 +12,152 @@
 #include "MG_Util/Converters/MGToStr/TextureEnumConverter.h"
 
 namespace MobileGL::MG_Backend::DirectVulkan {
+    static Bool IsCubeMapFaceUploadTarget(TextureUploadTarget target) {
+        return target >= TextureUploadTarget::CubeMapPositiveX &&
+               target <= TextureUploadTarget::CubeMapNegativeZ;
+    }
+
+    static Bool PendingClearMatchesTextureIdentity(const PendingClearKey& key, const TextureIdentity& identity) {
+        return key.texture == identity.texture && key.textureLifetimeId == identity.lifetimeId;
+    }
+
+    static Uint32 ResolveAttachmentBaseArrayLayer(TextureUploadTarget target) {
+        if (!IsCubeMapFaceUploadTarget(target)) {
+            return 0;
+        }
+        return static_cast<Uint32>(target) - static_cast<Uint32>(TextureUploadTarget::CubeMapPositiveX);
+    }
+
+    static Uint32 ResolveAttachmentBaseArrayLayer(
+        const MG_State::GLState::FramebufferAttachmentObject& attachment) {
+        if (attachment.IsLayered()) {
+            return 0;
+        }
+        const TextureUploadTarget uploadTarget = attachment.GetTextureUploadTarget();
+        if (!IsCubeMapFaceUploadTarget(uploadTarget)) {
+            return static_cast<Uint32>(std::max(attachment.GetTextureLayer(), 0));
+        }
+        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;
+    }
+
+    static const MG_State::GLState::FramebufferAttachmentObject* GetClearableAttachment(
+        const MG_State::GLState::FramebufferObject& drawFbo, FramebufferAttachmentType attachmentType) {
+        if (attachmentType == FramebufferAttachmentType::None) {
+            return nullptr;
+        }
+
+        const auto& attachment = drawFbo.GetAttachment(attachmentType);
+        if (!attachment.IsTexture() || attachment.IsRenderbuffer()) {
+            return nullptr;
+        }
+
+        return &attachment;
+    }
+
+    PendingClearKey VkClearManager::MakePendingClearKey(MG_State::GLState::ITextureObject* texture, Uint32 mipLevel,
+                                                        Uint32 baseArrayLayer, Uint32 layerCount) {
+        return PendingClearKey {
+            .texture = texture,
+            .textureLifetimeId = texture ? texture->GetLifetimeId() : 0,
+            .mipLevel = mipLevel,
+            .baseArrayLayer = baseArrayLayer,
+            .layerCount = layerCount,
+        };
+    }
+
+    PendingClearKey VkClearManager::MakePendingClearKey(
+        const MG_State::GLState::FramebufferAttachmentObject& attachment) {
+        MOBILEGL_ASSERT(attachment.IsTexture() && !attachment.IsRenderbuffer(),
+                        "MakePendingClearKey requires a texture framebuffer attachment");
+        auto* texture = attachment.GetTexture().get();
+        MOBILEGL_ASSERT(texture != nullptr, "MakePendingClearKey: texture attachment resolved to null");
+        const Uint32 mipLevel = static_cast<Uint32>(std::max(attachment.GetTextureLevel(), 0));
+        const Uint32 baseArrayLayer = ResolveAttachmentBaseArrayLayer(attachment);
+        const Uint32 layerCount = ResolveAttachmentLayerCount(attachment);
+        return MakePendingClearKey(texture, mipLevel, baseArrayLayer, layerCount);
+    }
+
     Bool VkClearManager::Initialize() {
         return true;
     }
 
     void VkClearManager::Shutdown() {
+        const std::lock_guard<std::mutex> lock(m_mutex);
+        m_pendingClears.clear();
+        m_aliveObjects.clear();
+    }
 
+    TextureIdentity VkClearManager::MakeTextureIdentity(MG_State::GLState::ITextureObject* texture) {
+        return TextureIdentity {
+            .texture = texture,
+            .lifetimeId = texture ? texture->GetLifetimeId() : 0,
+        };
+    }
+
+    void VkClearManager::MergeClearPayload(ClearAttachmentPayload& dst, const ClearAttachmentPayload& src) {
+        dst.mask |= src.mask;
+        if ((src.mask & GL_COLOR_BUFFER_BIT) != 0) {
+            dst.color = src.color;
+        }
+        if ((src.mask & GL_DEPTH_BUFFER_BIT) != 0) {
+            dst.depth = src.depth;
+        }
+        if ((src.mask & GL_STENCIL_BUFFER_BIT) != 0) {
+            dst.stencil = src.stencil;
+        }
+    }
+
+    void VkClearManager::ErasePendingClearsForTextureLocked(const TextureIdentity& identity) {
+        Vector<PendingClearKey> keysToErase;
+        keysToErase.reserve(m_pendingClears.size());
+        for (auto it = m_pendingClears.begin(); it != m_pendingClears.end(); ++it) {
+            if (PendingClearMatchesTextureIdentity(it->first, identity)) {
+                keysToErase.emplace_back(it->first);
+            }
+        }
+        for (const auto& key : keysToErase) {
+            m_pendingClears.erase(key);
+        }
+        m_aliveObjects.erase(identity);
+    }
+
+    Bool VkClearManager::LockTextureIdentityLocked(const TextureIdentity& identity,
+                                                   SharedPtr<MG_State::GLState::ITextureObject>& outTexture) {
+        outTexture.reset();
+        if (identity.texture == nullptr) {
+            return false;
+        }
+
+        auto aliveIt = m_aliveObjects.find(identity);
+        if (aliveIt == m_aliveObjects.end()) {
+            ErasePendingClearsForTextureLocked(identity);
+            return false;
+        }
+
+        outTexture = aliveIt->second.lock();
+        if (!outTexture || outTexture.get() != identity.texture || outTexture->GetLifetimeId() != identity.lifetimeId) {
+            ErasePendingClearsForTextureLocked(identity);
+            outTexture.reset();
+            return false;
+        }
+
+        return true;
+    }
+
+    Bool VkClearManager::LockTextureLocked(const PendingClearKey& key,
+                                           SharedPtr<MG_State::GLState::ITextureObject>& outTexture) {
+        return LockTextureIdentityLocked(TextureIdentity{
+            .texture = key.texture,
+            .lifetimeId = key.textureLifetimeId,
+        }, outTexture);
     }
 
     void VkClearManager::QueueClear(GLbitfield mask, const ClearFramebufferPayload& clearPayload,
@@ -26,93 +166,237 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             auto& drawbufs = drawFbo.GetDrawBuffers();
             // This should automatically work on default & offscreen FBO
             for (auto drawbuf: drawbufs) {
-                if (drawbuf == FramebufferAttachmentType::None ||
-                    drawFbo.GetAttachment(drawbuf).IsRenderbuffer())
+                const auto* attachment = GetClearableAttachment(drawFbo, drawbuf);
+                if (!attachment) {
                     continue;
+                }
 
                 QueueClear({
-                    .color = clearPayload.color,
-                    .attachmentType = drawbuf
-                }, drawFbo.GetAttachment(drawbuf).GetTexture());
+                    .mask = GL_COLOR_BUFFER_BIT,
+                    .color = clearPayload.color
+                }, *attachment);
+
                 MGLOG_D("%s: %s (texture %d) - color = (%.2f, %.2f, %.2f, %.2f)", __func__,
                     MG_Util::ConvertFramebufferAttachmentTypeToString(drawbuf).c_str(),
-                    drawFbo.GetAttachment(drawbuf).GetTexture()->GetExternalIndex(),
+                    attachment->GetTexture()->GetExternalIndex(),
                     clearPayload.color[0], clearPayload.color[1], clearPayload.color[2], clearPayload.color[3]);
             }
         }
 
-        if (mask & GL_DEPTH_BUFFER_BIT &&
-            !drawFbo.GetAttachment(FramebufferAttachmentType::Depth).IsRenderbuffer()) {
-            QueueClear({
-                .depth = clearPayload.depth,
-                .attachmentType = FramebufferAttachmentType::Depth,
-            }, drawFbo.GetAttachment(FramebufferAttachmentType::Depth).GetTexture());
-            MGLOG_D("%s: Depth (texture %d) - depth = (%.2f)", __func__,
-                drawFbo.GetAttachment(FramebufferAttachmentType::Depth).GetTexture()->GetExternalIndex(), clearPayload.depth);
+        if (mask & GL_DEPTH_BUFFER_BIT) {
+            const auto* attachment = GetClearableAttachment(drawFbo, FramebufferAttachmentType::Depth);
+            if (attachment) {
+                QueueClear({
+                    .mask = GL_DEPTH_BUFFER_BIT,
+                    .depth = clearPayload.depth,
+                }, *attachment);
+
+                MGLOG_D("%s: Depth (texture %d) - depth = (%.2f)", __func__,
+                    attachment->GetTexture()->GetExternalIndex(), clearPayload.depth);
+            }
         }
 
-        if (mask & GL_STENCIL_BUFFER_BIT &&
-            !drawFbo.GetAttachment(FramebufferAttachmentType::Stencil).IsRenderbuffer()) {
-            QueueClear({
-                .stencil = clearPayload.stencil,
-                .attachmentType = FramebufferAttachmentType::Stencil,
-            }, drawFbo.GetAttachment(FramebufferAttachmentType::Stencil).GetTexture());
-            MGLOG_D("%s: Stencil (texture %d) - stencil = (%u)", __func__,
-                drawFbo.GetAttachment(FramebufferAttachmentType::Stencil).GetTexture()->GetExternalIndex(), clearPayload.stencil);
+        if (mask & GL_STENCIL_BUFFER_BIT) {
+            const auto* attachment = GetClearableAttachment(drawFbo, FramebufferAttachmentType::Stencil);
+            if (attachment) {
+                QueueClear({
+                    .mask = GL_STENCIL_BUFFER_BIT,
+                    .stencil = clearPayload.stencil,
+                }, *attachment);
+
+                MGLOG_D("%s: Stencil (texture %d) - stencil = (%u)", __func__,
+                    attachment->GetTexture()->GetExternalIndex(), clearPayload.stencil);
+            }
         }
     }
 
     void VkClearManager::QueueClear(const ClearAttachmentPayload& clearPayload,
                                     const SharedPtr<MG_State::GLState::ITextureObject>& texture) {
-        WeakPtr<MG_State::GLState::ITextureObject> weakTexturePtr = texture;
-        if (weakTexturePtr.expired())
+        if (clearPayload.mask == 0 || !texture) {
             return;
-        auto* pTexture = weakTexturePtr.lock().get();
-        m_aliveObjects[pTexture] = weakTexturePtr;
-        m_pendingClears[pTexture] = clearPayload;
+        }
+
+        const PendingClearKey key = MakePendingClearKey(texture.get());
+        const std::lock_guard<std::mutex> lock(m_mutex);
+        m_aliveObjects[MakeTextureIdentity(texture.get())] = texture;
+        auto& pending = m_pendingClears[key];
+        MergeClearPayload(pending, clearPayload);
+    }
+
+    void VkClearManager::QueueClear(const ClearAttachmentPayload& clearPayload,
+                                    const MG_State::GLState::FramebufferAttachmentObject& attachment) {
+        if (clearPayload.mask == 0 || !attachment.IsTexture() || attachment.IsRenderbuffer()) {
+            return;
+        }
+        const auto texture = attachment.GetTexture();
+        if (!texture) {
+            return;
+        }
+
+        const PendingClearKey key = MakePendingClearKey(attachment);
+        const std::lock_guard<std::mutex> lock(m_mutex);
+        m_aliveObjects[MakeTextureIdentity(texture.get())] = texture;
+        auto& pending = m_pendingClears[key];
+        MergeClearPayload(pending, clearPayload);
     }
 
     Bool VkClearManager::HasPendingClear(MG_State::GLState::ITextureObject* texture) {
-        return m_pendingClears.find(texture) != m_pendingClears.end();
-    }
-
-    Bool VkClearManager::GetPendingClear(MG_State::GLState::ITextureObject* texture, ClearAttachmentPayload& outPayload) {
-        if (m_aliveObjects.find(texture) == m_aliveObjects.end() ||
-            m_pendingClears.find(texture) == m_pendingClears.end()) {
-            MGLOG_D("%s: Failed getting pending clear for texture %d", __func__, texture->GetExternalIndex());
+        if (texture == nullptr) {
             return false;
         }
 
-        outPayload = m_pendingClears[texture];
-        MGLOG_D("%s: Got pending clear for texture %d (%s), clear value: color = (%.2f, %.2f, %.2f, %.2f), depth = (%.2f), stencil = (%u)", __func__,
-            texture->GetExternalIndex(),
-            MG_Util::ConvertTextureInternalFormatToString(texture->GetFormat()).c_str(),
+        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) {
+            if (it->first.texture == texture && it->first.textureLifetimeId == lifetimeId) {
+                SharedPtr<MG_State::GLState::ITextureObject> liveTexture;
+                return LockTextureLocked(it->first, liveTexture);
+            }
+        }
+        return false;
+    }
+
+    Bool VkClearManager::HasPendingClear(const PendingClearKey& key) {
+        if (key.texture == nullptr) {
+            return false;
+        }
+
+        const std::lock_guard<std::mutex> lock(m_mutex);
+        if (m_pendingClears.find(key) == m_pendingClears.end()) {
+            return false;
+        }
+
+        SharedPtr<MG_State::GLState::ITextureObject> liveTexture;
+        return LockTextureLocked(key, liveTexture);
+    }
+
+    Bool VkClearManager::HasPendingClear(const MG_State::GLState::FramebufferAttachmentObject& attachment) {
+        if (!attachment.IsTexture() || attachment.IsRenderbuffer() || !attachment.GetTexture()) {
+            return false;
+        }
+        return HasPendingClear(MakePendingClearKey(attachment));
+    }
+
+    Bool VkClearManager::GetPendingClear(const PendingClearKey& key, ClearAttachmentPayload& outPayload) {
+        SharedPtr<MG_State::GLState::ITextureObject> liveTexture;
+        return GetPendingClear(key, outPayload, liveTexture);
+    }
+
+    Bool VkClearManager::GetPendingClear(const PendingClearKey& key, ClearAttachmentPayload& outPayload,
+                                         SharedPtr<MG_State::GLState::ITextureObject>& outTexture) {
+        if (key.texture == nullptr) {
+            return false;
+        }
+
+        const std::lock_guard<std::mutex> lock(m_mutex);
+        if (!LockTextureLocked(key, outTexture)) {
+            return false;
+        }
+        auto it = m_pendingClears.find(key);
+        if (it == m_pendingClears.end()) {
+            outTexture.reset();
+            return false;
+        }
+
+        outPayload = it->second;
+        MGLOG_D("%s: Got pending clear for texture@%p lifetime=%llu, mip=%u layer=%u count=%u mask=0x%x clear value: color = (%.2f, %.2f, %.2f, %.2f), depth = (%.2f), stencil = (%u)", __func__,
+            static_cast<void*>(key.texture),
+            static_cast<unsigned long long>(key.textureLifetimeId),
+            key.mipLevel, key.baseArrayLayer, key.layerCount,
+            static_cast<Uint32>(outPayload.mask),
             outPayload.color[0], outPayload.color[1], outPayload.color[2], outPayload.color[3],
             outPayload.depth,
             outPayload.stencil);
         return true;
     }
 
+    Bool VkClearManager::GetPendingClear(const MG_State::GLState::FramebufferAttachmentObject& attachment,
+                                         ClearAttachmentPayload& outPayload) {
+        if (!attachment.IsTexture() || attachment.IsRenderbuffer() || !attachment.GetTexture()) {
+            MGLOG_D("%s: Failed getting pending clear for non-texture framebuffer attachment", __func__);
+            return false;
+        }
+        return GetPendingClear(MakePendingClearKey(attachment), outPayload);
+    }
+
+    Bool VkClearManager::GetPendingClears(MG_State::GLState::ITextureObject* texture,
+                                          Vector<PendingClearEntry>& outEntries) {
+        outEntries.clear();
+        if (texture == nullptr) {
+            return false;
+        }
+
+        const Uint64 lifetimeId = texture->GetLifetimeId();
+        const std::lock_guard<std::mutex> lock(m_mutex);
+        SharedPtr<MG_State::GLState::ITextureObject> liveTexture;
+        if (!LockTextureIdentityLocked(MakeTextureIdentity(texture), liveTexture)) {
+            return false;
+        }
+        for (auto it = m_pendingClears.begin(); it != m_pendingClears.end(); ++it) {
+            if (it->first.texture == texture && it->first.textureLifetimeId == lifetimeId) {
+                outEntries.emplace_back(PendingClearEntry{.key = it->first, .payload = it->second});
+            }
+        }
+        return !outEntries.empty();
+    }
+
     void VkClearManager::PopPendingClear(MG_State::GLState::ITextureObject* texture) {
-        MGLOG_D("%s: Pop pending clear for texture %d", __func__, texture->GetExternalIndex());
-        m_aliveObjects.erase(texture);
-        m_pendingClears.erase(texture);
+        if (texture == nullptr) {
+            return;
+        }
+
+        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);
+        ErasePendingClearsForTextureLocked(identity);
+    }
+
+    void VkClearManager::PopPendingClear(const PendingClearKey& key) {
+        if (key.texture == nullptr) {
+            return;
+        }
+
+        {
+            const std::lock_guard<std::mutex> lock(m_mutex);
+            auto it = m_pendingClears.find(key);
+            if (it != m_pendingClears.end()) {
+                m_pendingClears.erase(it);
+            }
+        }
+
+        MGLOG_D("%s: Pop pending clear for texture@%p lifetime=%llu mip=%u layer=%u count=%u", __func__,
+                static_cast<void*>(key.texture), static_cast<unsigned long long>(key.textureLifetimeId),
+                key.mipLevel, key.baseArrayLayer, key.layerCount);
+    }
+
+    void VkClearManager::PopPendingClear(const MG_State::GLState::FramebufferAttachmentObject& attachment) {
+        if (!attachment.IsTexture() || attachment.IsRenderbuffer() || !attachment.GetTexture()) {
+            return;
+        }
+        PopPendingClear(MakePendingClearKey(attachment));
     }
 
     SizeT VkClearManager::CollectGarbage() {
+        const std::lock_guard<std::mutex> lock(m_mutex);
         m_gcCounter++;
         if (m_gcCounter != 0) {
             return 0;
         }
 
-        SizeT count = 0;
-        for (const auto& [raw, weak]: m_aliveObjects) {
-            if (weak.expired()) {
-                count++;
-                m_pendingClears.erase(raw);
-                m_aliveObjects.erase(raw);
+        Vector<TextureIdentity> expiredTextures;
+        expiredTextures.reserve(m_aliveObjects.size());
+        for (auto it = m_aliveObjects.begin(); it != m_aliveObjects.end(); ++it) {
+            if (it->second.expired()) {
+                expiredTextures.emplace_back(it->first);
             }
         }
-        return count;
+        if (expiredTextures.empty()) {
+            return 0;
+        }
+
+        for (const auto& identity : expiredTextures) {
+            ErasePendingClearsForTextureLocked(identity);
+        }
+        return expiredTextures.size();
     }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkClearManager.h

@@ -14,6 +14,7 @@
 #include "MG_Util/Math/VectorTypes.h"
 
 #include <Includes.h>
+#include <unordered_map>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
     struct ClearFramebufferPayload {
@@ -23,16 +24,70 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     };
 
     struct ClearAttachmentPayload {
-        union {
-            FloatVec4 color;
-            Float depth{};
-            Uint32 stencil;
-        };
-        FramebufferAttachmentType attachmentType = FramebufferAttachmentType::Color0;
+        GLbitfield mask = 0;
+        FloatVec4 color = FloatVec4(0.0f, 0.0f, 0.0f, 0.0f);
+        Float depth = 1.0f;
+        Uint32 stencil = 0;
+    };
+
+    struct PendingClearKey {
+        MG_State::GLState::ITextureObject* texture = nullptr;
+        Uint64 textureLifetimeId = 0;
+        Uint32 mipLevel = 0;
+        Uint32 baseArrayLayer = 0;
+        Uint32 layerCount = 1;
+
+        Bool operator==(const PendingClearKey& other) const {
+            return texture == other.texture && textureLifetimeId == other.textureLifetimeId &&
+                   mipLevel == other.mipLevel &&
+                   baseArrayLayer == other.baseArrayLayer && layerCount == other.layerCount;
+        }
+    };
+
+    struct TextureIdentity {
+        MG_State::GLState::ITextureObject* texture = nullptr;
+        Uint64 lifetimeId = 0;
+
+        Bool operator==(const TextureIdentity& other) const {
+            return texture == other.texture && lifetimeId == other.lifetimeId;
+        }
+    };
+
+    struct PendingClearEntry {
+        PendingClearKey key{};
+        ClearAttachmentPayload payload{};
+    };
+
+    struct PendingClearKeyHash {
+        SizeT operator()(const PendingClearKey& key) const {
+            const SizeT textureHash = std::hash<MG_State::GLState::ITextureObject*>{}(key.texture);
+            const SizeT textureLifetimeHash = std::hash<Uint64>{}(key.textureLifetimeId);
+            const SizeT mipHash = std::hash<Uint32>{}(key.mipLevel);
+            const SizeT layerHash = std::hash<Uint32>{}(key.baseArrayLayer);
+            const SizeT layerCountHash = std::hash<Uint32>{}(key.layerCount);
+            SizeT hash = textureHash;
+            hash ^= textureLifetimeHash + 0x9e3779b9u + (hash << 6) + (hash >> 2);
+            hash ^= mipHash + 0x9e3779b9u + (hash << 6) + (hash >> 2);
+            hash ^= layerHash + 0x9e3779b9u + (hash << 6) + (hash >> 2);
+            hash ^= layerCountHash + 0x9e3779b9u + (hash << 6) + (hash >> 2);
+            return hash;
+        }
+    };
+
+    struct TextureIdentityHash {
+        SizeT operator()(const TextureIdentity& key) const {
+            SizeT hash = std::hash<MG_State::GLState::ITextureObject*>{}(key.texture);
+            hash ^= std::hash<Uint64>{}(key.lifetimeId) + 0x9e3779b9u + (hash << 6) + (hash >> 2);
+            return hash;
+        }
     };
 
     class VkClearManager {
     public:
+        static PendingClearKey MakePendingClearKey(const MG_State::GLState::FramebufferAttachmentObject& attachment);
+        static PendingClearKey MakePendingClearKey(MG_State::GLState::ITextureObject* texture, Uint32 mipLevel = 0,
+                                                   Uint32 baseArrayLayer = 0, Uint32 layerCount = 1);
+
         Bool Initialize();
         void Shutdown();
 
@@ -40,13 +95,33 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         void QueueClear(
             const ClearAttachmentPayload& clearPayload,
             const SharedPtr<MG_State::GLState::ITextureObject>& texture);
+        void QueueClear(const ClearAttachmentPayload& clearPayload,
+                        const MG_State::GLState::FramebufferAttachmentObject& attachment);
         Bool HasPendingClear(MG_State::GLState::ITextureObject* texture);
-        Bool GetPendingClear(MG_State::GLState::ITextureObject* texture, ClearAttachmentPayload& outPayload);
+        Bool HasPendingClear(const PendingClearKey& key);
+        Bool HasPendingClear(const MG_State::GLState::FramebufferAttachmentObject& attachment);
+        Bool GetPendingClear(const PendingClearKey& key, ClearAttachmentPayload& outPayload);
+        Bool GetPendingClear(const PendingClearKey& key, ClearAttachmentPayload& outPayload,
+                             SharedPtr<MG_State::GLState::ITextureObject>& outTexture);
+        Bool GetPendingClear(const MG_State::GLState::FramebufferAttachmentObject& attachment,
+                             ClearAttachmentPayload& outPayload);
+        Bool GetPendingClears(MG_State::GLState::ITextureObject* texture, Vector<PendingClearEntry>& outEntries);
         void PopPendingClear(MG_State::GLState::ITextureObject* texture);
+        void PopPendingClear(const PendingClearKey& key);
+        void PopPendingClear(const MG_State::GLState::FramebufferAttachmentObject& attachment);
         SizeT CollectGarbage();
     private:
+        static TextureIdentity MakeTextureIdentity(MG_State::GLState::ITextureObject* texture);
+        static void MergeClearPayload(ClearAttachmentPayload& dst, const ClearAttachmentPayload& src);
+        void ErasePendingClearsForTextureLocked(const TextureIdentity& identity);
+        Bool LockTextureIdentityLocked(const TextureIdentity& identity,
+                                      SharedPtr<MG_State::GLState::ITextureObject>& outTexture);
+        Bool LockTextureLocked(const PendingClearKey& key,
+                               SharedPtr<MG_State::GLState::ITextureObject>& outTexture);
+
         Uint8 m_gcCounter = 0;
-        UnorderedMap<MG_State::GLState::ITextureObject*, ClearAttachmentPayload> m_pendingClears;
-        UnorderedMap<MG_State::GLState::ITextureObject*, WeakPtr<MG_State::GLState::ITextureObject>> m_aliveObjects;
+        mutable std::mutex m_mutex;
+        std::unordered_map<PendingClearKey, ClearAttachmentPayload, PendingClearKeyHash> m_pendingClears;
+        std::unordered_map<TextureIdentity, WeakPtr<MG_State::GLState::ITextureObject>, TextureIdentityHash> m_aliveObjects;
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkRenderPassManager.h

@@ -16,26 +16,49 @@
 #include "MG_State/GLState/FramebufferState/FramebufferObject.h"
 
 #include <Includes.h>
+#include <vk_mem_alloc.h>
 
 namespace MobileGL::MG_Backend::DirectVulkan {
     enum class TrackedAttachmentTarget : Uint8 {
         Texture,
+        Renderbuffer,
         SwapchainColor,
         SwapchainDepthStencil
     };
 
     struct PendingClearAttachmentInfo {
+        // Index into the render pass attachment descriptions (VkRenderPassBeginInfo::pClearValues space).
         Uint32 attachmentIndex = 0;
-        MG_State::GLState::ITextureObject* texture = nullptr;
+        // Index into the subpass pColorAttachments (VkClearAttachment::colorAttachment space) — the GL
+        // draw-buffer slot. Differs from attachmentIndex when earlier slots are GL_NONE/incomplete.
+        // Only meaningful for color clears.
+        Uint32 colorAttachmentSlot = 0;
+        PendingClearKey key{};
+        MG_State::GLState::RenderbufferObject* renderbuffer = nullptr;
+        Bool hasInlinePayload = false;
+        ClearAttachmentPayload inlinePayload{};
     };
 
     struct TrackedAttachmentLayoutInfo {
         TrackedAttachmentTarget target = TrackedAttachmentTarget::Texture;
-        MG_State::GLState::ITextureObject* texture = nullptr;
+        WeakPtr<MG_State::GLState::ITextureObject> texture;
+        WeakPtr<MG_State::GLState::RenderbufferObject> renderbuffer;
+        Uint32 textureMipLevel = 0;
         Uint32 swapchainImageIndex = 0;
         VkImageLayout finalLayout = VK_IMAGE_LAYOUT_UNDEFINED;
     };
 
+    struct DepthStencilAttachmentLoadInfo {
+        VkAttachmentLoadOp depthLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
+        VkAttachmentLoadOp stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
+        VkImageLayout initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
+    };
+
+    DepthStencilAttachmentLoadInfo ResolveDepthStencilAttachmentLoadInfo(
+        VkImageLayout trackedLayout, Bool clearDepth, Bool clearStencil);
+    IntVec2 ResolveRenderPassFramebufferExtent(Bool isDefaultFbo, const TextureSize& attachmentExtent,
+                                               VkExtent2D swapchainExtent);
+
     struct RenderPassEntry {
         static inline VkDevice s_device;
         static inline Vector<VkTextureManager::TextureResource*> s_textureResourcesScratch;
@@ -46,8 +69,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Vector<PendingClearAttachmentInfo> pendingClearAttachments;
         Vector<TrackedAttachmentLayoutInfo> trackedAttachmentLayouts;
         Uint32 attachmentCount = 0;
+        Uint32 colorAttachmentCount = 0;
+        Bool hasDepthStencilAttachment = false;
+        VkSampleCountFlagBits sampleCount = VK_SAMPLE_COUNT_1_BIT;
         IntVec2 extent = {0, 0};
-        Uint32 subpass = 0;
+        // VkFramebufferCreateInfo::layers of the entry's framebuffer (>1 for layered GL attachments).
+        Uint32 layers = 1;
+        // Frame counter value of the last GetOrCreateRenderPass hit; drives cache eviction.
+        Uint64 lastUsedFrame = 0;
 
         RenderPassEntry() = default;
         RenderPassEntry(const RenderPassEntry&) = delete;
@@ -59,8 +88,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             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(subpass, that.subpass);
+            std::swap(layers, that.layers);
+            std::swap(lastUsedFrame, that.lastUsedFrame);
         }
         RenderPassEntry(
             Uint64 hash,
@@ -70,7 +103,10 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             const Vector<PendingClearAttachmentInfo>& pendingClearAttachments,
             const Vector<TrackedAttachmentLayoutInfo>& trackedAttachmentLayouts,
             Uint32 attachmentCount,
-            IntVec2 extent, int subpass):
+            Uint32 colorAttachmentCount,
+            Bool hasDepthStencilAttachment,
+            VkSampleCountFlagBits sampleCount,
+            IntVec2 extent, Uint32 layers):
             hash(hash),
             renderPass(renderpass),
             framebuffer(framebuffer),
@@ -78,8 +114,11 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             pendingClearAttachments(Move(pendingClearAttachments)),
             trackedAttachmentLayouts(Move(trackedAttachmentLayouts)),
             attachmentCount(attachmentCount),
+            colorAttachmentCount(colorAttachmentCount),
+            hasDepthStencilAttachment(hasDepthStencilAttachment),
+            sampleCount(sampleCount),
             extent(extent),
-            subpass(subpass)
+            layers(layers)
         {}
 
         ~RenderPassEntry() {
@@ -119,8 +158,8 @@ namespace MobileGL::MG_Backend::DirectVulkan {
     public:
         using HashType = Uint64;
         VkRenderPassManager(VkDevice device,
-            const VulkanRendererConfig& config, VkClearManager& clearManager, VkTextureManager& textureManager,
-            SwapchainObject& swapchainObject);
+            VkPhysicalDevice physicalDevice, VmaAllocator allocator, const VulkanRendererConfig& config,
+            VkClearManager& clearManager, VkTextureManager& textureManager, SwapchainObject& swapchainObject);
         ~VkRenderPassManager();
 
         Bool Initialize();
@@ -129,23 +168,92 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         HashType ComputeHash(
             const MG_State::GLState::FramebufferObject& fbo,
             Uint32 swapchainImageIndex,
-            Bool includePendingClear = true) const;
+            Bool includePendingClear = true);
         RenderPassEntry& GetOrCreateRenderPass(const MG_State::GLState::FramebufferObject& fbo, Uint32 swapchainImageIndex);
+        void QueueRenderbufferClear(GLbitfield mask, const ClearFramebufferPayload& clearPayload,
+                                    const MG_State::GLState::FramebufferObject& drawFbo);
+        void QueueRenderbufferClear(const ClearAttachmentPayload& clearPayload,
+                                    const MG_State::GLState::FramebufferAttachmentObject& attachment);
+        void PopPendingRenderbufferClear(MG_State::GLState::RenderbufferObject* renderbuffer);
+        // Frame boundary hook: ages the render-pass cache and evicts long-unused
+        // entries (their command buffers retired many frames ago).
+        void OnPresent();
         static Bool BeginRenderPass(VkCommandBuffer commandBuffer, RenderPassEntry& renderPassEntry);
         static Bool EndRenderPass(VkCommandBuffer commandBuffer);
         static ActiveRenderPassInfo* GetActiveRenderPass();
     private:
         VkDevice m_device = VK_NULL_HANDLE;
+        VkPhysicalDevice m_physicalDevice = VK_NULL_HANDLE;
+        VmaAllocator m_allocator = nullptr;
         const VulkanRendererConfig& m_config;
         VkClearManager& m_clearManager;
         VkTextureManager& m_textureManager;
         SwapchainObject& m_swapchainObject;
         UnorderedMap<Uint64, RenderPassEntry> m_renderPasses;
+        // Monotonic frame counter (bumped in OnPresent) for render-pass cache aging.
+        Uint64 m_frameCounter = 0;
+
+        // 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;
+
+        // 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 /
+        // version change, swapchain rotation, any attachment image recreation (the two epochs),
+        // 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;
+        Uint16 m_rpFastFboVersion = 0;
+        Uint32 m_rpFastSwapchainIndex = 0;
+        Uint64 m_rpFastTexEpoch = 0;
+        Uint64 m_rpFastRbEpoch = 0;
+        Uint64 m_rpFastRenderPassHash = 0;
+
+    public:
+        struct RenderbufferResource {
+            WeakPtr<MG_State::GLState::RenderbufferObject> renderbuffer;
+            VkImage image = VK_NULL_HANDLE;
+            VmaAllocation allocation = nullptr;
+            VkImageView view = VK_NULL_HANDLE;
+            VkImageLayout layout = VK_IMAGE_LAYOUT_UNDEFINED;
+            VkFormat format = VK_FORMAT_UNDEFINED;
+            VkImageAspectFlags aspect = VK_IMAGE_ASPECT_NONE;
+            VkExtent2D extent = {0, 0};
+            VkSampleCountFlagBits sampleCount = VK_SAMPLE_COUNT_1_BIT;
+            TextureInternalFormat internalFormat = TextureInternalFormat::Unknown;
+            Int samples = 0;
+
+            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;
+        UnorderedMap<MG_State::GLState::RenderbufferObject*, PendingRenderbufferClear> m_pendingRenderbufferClears;
+
+        Bool HasPendingRenderbufferClear(
+            const MG_State::GLState::FramebufferAttachmentObject& attachment) const;
+        void CollectRenderbufferGarbage();
+
         static inline XXH64_state_t* m_hashState = XXH64_createState();
         static inline ActiveRenderPassInfo s_activeRenderPass{};
         static inline Bool s_hasActiveRenderPass = false;
         static inline VkClearManager* s_clearManager = nullptr;
         static inline VkTextureManager* s_textureManager = nullptr;
         static inline SwapchainObject* s_swapchainObject = nullptr;
+        static inline VkRenderPassManager* s_renderPassManager = nullptr;
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkSamplerManager.cpp

@@ -10,17 +10,74 @@
 
 #include "MG_State/GLState/Core.h"
 
+#include <algorithm>
+#include <cmath>
+
 namespace MobileGL::MG_Backend::DirectVulkan {
+    namespace {
+        Bool UsesBorderColor(const MG_State::GLState::SamplerObject& sampler) {
+            return sampler.GetWrapS() == SamplerWrapMode::ClampToBorder ||
+                   sampler.GetWrapT() == SamplerWrapMode::ClampToBorder ||
+                   sampler.GetWrapR() == SamplerWrapMode::ClampToBorder;
+        }
+
+        Bool IsDepthTextureFormat(TextureInternalFormat format) {
+            switch (format) {
+            case TextureInternalFormat::DepthComponent:
+            case TextureInternalFormat::DepthComponent16:
+            case TextureInternalFormat::DepthComponent24:
+            case TextureInternalFormat::DepthComponent32:
+            case TextureInternalFormat::DepthComponent32F:
+            case TextureInternalFormat::Depth24Stencil8:
+            case TextureInternalFormat::Depth32FStencil8:
+            case TextureInternalFormat::DepthStencil:
+                return true;
+            default:
+                return false;
+            }
+        }
+
+        Bool NearlyEqual(Float lhs, Float rhs) {
+            return std::fabs(lhs - rhs) <= 1e-6f;
+        }
+
+        Float ResolveEffectiveMaxLod(const MG_State::GLState::SamplerObject& sampler) {
+            if (sampler.GetMipmapMode() == SamplerMipmapMode::None) {
+                return 0.0f;
+            }
+            return sampler.GetMaxLod();
+        }
+
+        Float ResolveEffectiveMinLod(const MG_State::GLState::SamplerObject& sampler, Float effectiveMaxLod) {
+            return std::min(sampler.GetMinLod(), effectiveMaxLod);
+        }
+    } // namespace
+
     Bool VkSamplerManager::Initialize(const InitInfo& initInfo) {
         Shutdown();
 
         m_device = initInfo.device;
         m_config = initInfo.config;
+        m_samplerAnisotropySupported = initInfo.samplerAnisotropySupported;
+        m_maxSamplerAnisotropy = std::max(initInfo.maxSamplerAnisotropy, 1.0f);
         MOBILEGL_ASSERT(m_device != VK_NULL_HANDLE && m_config != nullptr,
                         "VkSamplerManager::Initialize failed: invalid initialization info");
         return true;
     }
 
+    Float VkSamplerManager::ResolveEffectiveMaxAnisotropy(const MG_State::GLState::SamplerObject& sampler,
+                                                           Bool forceNearestFiltering) const {
+        if (!m_samplerAnisotropySupported) return 1.0f;
+        if (forceNearestFiltering) return 1.0f;
+        // VUID-VkSamplerCreateInfo-anisotropyEnable-01071/01072: anisotropy requires both filters to
+        // be LINEAR and the value to sit within [1, limits.maxSamplerAnisotropy].
+        if (sampler.GetMinFilter() != SamplerFilterMode::Linear ||
+            sampler.GetMagFilter() != SamplerFilterMode::Linear) {
+            return 1.0f;
+        }
+        return std::clamp(sampler.GetMaxAnisotropy(), 1.0f, m_maxSamplerAnisotropy);
+    }
+
     void VkSamplerManager::Shutdown() {
         for (auto& [_, sampler] : m_samplers) {
             if (m_device != VK_NULL_HANDLE && sampler.handle != VK_NULL_HANDLE) {
@@ -34,10 +91,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         m_config = nullptr;
     }
 
-    Uint64 VkSamplerManager::BuildSamplerKey(const MG_State::GLState::SamplerObject& sampler) const {
+    Uint64 VkSamplerManager::BuildSamplerKey(const MG_State::GLState::SamplerObject& sampler,
+                                             const MG_State::GLState::ITextureObject& texture,
+                                             Bool forceNearestFiltering) 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)));
+
         const auto minFilter = sampler.GetMinFilter();
         XXHASH_VERIFY(XXH64_update(m_hashState, &minFilter, sizeof(minFilter)));
         const auto magFilter = sampler.GetMagFilter();
@@ -50,21 +111,30 @@ 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 minLod = sampler.GetMinLod();
+        const auto maxLod = ResolveEffectiveMaxLod(sampler);
+        const auto minLod = ResolveEffectiveMinLod(sampler, maxLod);
         XXHASH_VERIFY(XXH64_update(m_hashState, &minLod, sizeof(minLod)));
-        const auto maxLod = sampler.GetMaxLod();
         XXHASH_VERIFY(XXH64_update(m_hashState, &maxLod, sizeof(maxLod)));
         const auto lodBias = sampler.GetLodBias();
         XXHASH_VERIFY(XXH64_update(m_hashState, &lodBias, sizeof(lodBias)));
+        // The RESOLVED value, not the GL request: samplers that only differ in an anisotropy Vulkan
+        // will not apply (NEAREST filtering, or requests past the device limit) must still share one
+        // VkSampler, while two samplers that really do differ must not collide onto the first one's.
+        const auto maxAnisotropy = ResolveEffectiveMaxAnisotropy(sampler, forceNearestFiltering);
+        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 = sampler.GetSamplerCompareFunc();
+        const auto compareFunc = ResolveCompareFunc(sampler, texture);
         XXHASH_VERIFY(XXH64_update(m_hashState, &compareFunc, sizeof(compareFunc)));
+        const auto borderColor = ResolveVkBorderColor(sampler, texture);
+        XXHASH_VERIFY(XXH64_update(m_hashState, &borderColor, sizeof(borderColor)));
         return XXH64_digest(m_hashState);
     }
 
-    VkSampler VkSamplerManager::GetOrCreateSampler(const MG_State::GLState::SamplerObject& sampler) {
-        const Uint64 key = BuildSamplerKey(sampler);
+    VkSampler VkSamplerManager::GetOrCreateSampler(const MG_State::GLState::SamplerObject& sampler,
+                                                   const MG_State::GLState::ITextureObject& texture,
+                                                   Bool forceNearestFiltering) {
+        const Uint64 key = BuildSamplerKey(sampler, texture, forceNearestFiltering);
         auto it = m_samplers.find(key);
         if (it != m_samplers.end()) {
             return it->second.handle;
@@ -72,20 +142,24 @@ namespace MobileGL::MG_Backend::DirectVulkan {
 
         VkSamplerCreateInfo samplerInfo{};
         samplerInfo.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO;
-        samplerInfo.magFilter = ToVkFilter(sampler.GetMagFilter());
-        samplerInfo.minFilter = ToVkFilter(sampler.GetMinFilter());
-        samplerInfo.mipmapMode = ToVkMipmapMode(sampler.GetMipmapMode());
+        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.addressModeU = ToVkAddressMode(sampler.GetWrapS());
         samplerInfo.addressModeV = ToVkAddressMode(sampler.GetWrapT());
         samplerInfo.addressModeW = ToVkAddressMode(sampler.GetWrapR());
         samplerInfo.mipLodBias = sampler.GetLodBias();
-        samplerInfo.anisotropyEnable = VK_FALSE;
-        samplerInfo.maxAnisotropy = 1.0f;
+        // Must use the same resolver as BuildSamplerKey - a divergence would either collide two
+        // different samplers or silently create duplicates.
+        const Float maxAnisotropy = ResolveEffectiveMaxAnisotropy(sampler, forceNearestFiltering);
+        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(sampler.GetSamplerCompareFunc());
-        samplerInfo.minLod = sampler.GetMinLod();
-        samplerInfo.maxLod = sampler.GetMaxLod();
-        samplerInfo.borderColor = VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+        samplerInfo.compareOp = ToVkCompareOp(ResolveCompareFunc(sampler, texture));
+        samplerInfo.maxLod = ResolveEffectiveMaxLod(sampler);
+        samplerInfo.minLod = ResolveEffectiveMinLod(sampler, samplerInfo.maxLod);
+        samplerInfo.borderColor = ResolveVkBorderColor(sampler, texture);
         samplerInfo.unnormalizedCoordinates = VK_FALSE;
 
         VkSampler vkSampler = VK_NULL_HANDLE;
@@ -153,4 +227,49 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return VK_COMPARE_OP_ALWAYS;
         }
     }
+
+    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) {
+        if (!UsesBorderColor(sampler)) {
+            return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+        }
+
+        const auto& borderColor = texture.GetBorderColor();
+        const Bool isDepthTexture = IsDepthTextureFormat(texture.GetFormat());
+
+        if (isDepthTexture) {
+            if (NearlyEqual(borderColor.x(), 1.0f)) {
+                return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
+            }
+            if (NearlyEqual(borderColor.x(), 0.0f)) {
+                return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+            }
+        }
+
+        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;
+        }
+        if (rgbZero && NearlyEqual(borderColor.w(), 1.0f)) {
+            return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+        }
+        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;
+        }
+
+        return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+    }
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkSamplerManager.h

@@ -15,6 +15,7 @@
 
 namespace MobileGL::MG_State::GLState {
 class SamplerObject;
+class ITextureObject;
 }
 
 namespace MobileGL::MG_Backend::DirectVulkan {
@@ -23,12 +24,18 @@ public:
     struct InitInfo {
         VkDevice device = VK_NULL_HANDLE;
         const VulkanRendererConfig* config = nullptr;
+        // The samplerAnisotropy device feature was requested and granted at vkCreateDevice.
+        Bool samplerAnisotropySupported = false;
+        // VkPhysicalDeviceLimits::maxSamplerAnisotropy.
+        Float maxSamplerAnisotropy = 1.0f;
     };
 
     Bool Initialize(const InitInfo& initInfo);
     void Shutdown();
 
-    VkSampler GetOrCreateSampler(const MG_State::GLState::SamplerObject& sampler);
+    VkSampler GetOrCreateSampler(const MG_State::GLState::SamplerObject& sampler,
+                                 const MG_State::GLState::ITextureObject& texture,
+                                 Bool forceNearestFiltering = false);
 
 private:
     struct SamplerCacheEntry {
@@ -37,14 +44,28 @@ private:
         Uint16 version = 0;
     };
 
-    Uint64 BuildSamplerKey(const MG_State::GLState::SamplerObject& sampler) const;
+    Uint64 BuildSamplerKey(const MG_State::GLState::SamplerObject& sampler,
+                           const MG_State::GLState::ITextureObject& texture,
+                           Bool forceNearestFiltering) 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);
+    // 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
+    // that) while Vulkan forbids anisotropyEnable there, so the GL value must never be forwarded raw.
+    Float ResolveEffectiveMaxAnisotropy(const MG_State::GLState::SamplerObject& sampler,
+                                        Bool forceNearestFiltering) const;
 
     VkDevice m_device = VK_NULL_HANDLE;
     const VulkanRendererConfig* m_config = nullptr;
+    Bool m_samplerAnisotropySupported = false;
+    Float m_maxSamplerAnisotropy = 1.0f;
     UnorderedMap<Uint64, SamplerCacheEntry> m_samplers;
     static inline XXH64_state_t* m_hashState = XXH64_createState();
 };

MobileGL/MG_Backend/DirectVulkan/Renderer/VkTextureManager.h

@@ -12,35 +12,159 @@
 #include <Includes.h>
 #include <MG_State/GLState/TextureState/TextureObject.h>
 #include <vk_mem_alloc.h>
+#include <unordered_map>
+#include <unordered_set>
 
 namespace MobileGL::MG_State::GLState {
 class ITextureObject;
 }
 
 namespace MobileGL::MG_Backend::DirectVulkan {
+enum class SamplerNumericDomain : Uint8;
+
 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; }
+
+    struct TextureIdentity {
+        MG_State::GLState::ITextureObject* texture = nullptr;
+        Uint64 lifetimeId = 0;
+
+        Bool operator==(const TextureIdentity& other) const {
+            return texture == other.texture && lifetimeId == other.lifetimeId;
+        }
+    };
+
+    struct TextureIdentityHash {
+        SizeT operator()(const TextureIdentity& key) const {
+            SizeT hash = std::hash<MG_State::GLState::ITextureObject*>{}(key.texture);
+            hash ^= std::hash<Uint64>{}(key.lifetimeId) + 0x9e3779b9u + (hash << 6) + (hash >> 2);
+            return hash;
+        }
+    };
+
     struct InitInfo {
         VkDevice device = VK_NULL_HANDLE;
         VkPhysicalDevice physicalDevice = VK_NULL_HANDLE;
         VmaAllocator allocator = nullptr;
         VkCommandPool commandPool = VK_NULL_HANDLE;
         VkQueue graphicsQueue = VK_NULL_HANDLE;
+        Uint32 frameCount = 0;
     };
 
     struct TextureResource {
+        struct AttachmentViewKey {
+            Uint32 mipLevel = 0;
+            Uint32 baseArrayLayer = 0;
+            Uint32 layerCount = 1;
+            VkImageViewType viewType = VK_IMAGE_VIEW_TYPE_2D;
+
+            Bool operator==(const AttachmentViewKey& other) const {
+                return mipLevel == other.mipLevel &&
+                       baseArrayLayer == other.baseArrayLayer &&
+                       layerCount == other.layerCount &&
+                       viewType == other.viewType;
+            }
+        };
+
+        struct AttachmentViewKeyHash {
+            SizeT operator()(const AttachmentViewKey& key) const {
+                SizeT hash = std::hash<Uint32>{}(key.mipLevel);
+                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);
+                return hash;
+            }
+        };
+
+        struct StorageImageViewKey {
+            Uint32 mipLevel = 0;
+            Uint32 baseArrayLayer = 0;
+            Uint32 layerCount = 1;
+            VkImageViewType viewType = VK_IMAGE_VIEW_TYPE_2D;
+            VkFormat format = VK_FORMAT_UNDEFINED;
+
+            Bool operator==(const StorageImageViewKey& other) const {
+                return mipLevel == other.mipLevel &&
+                       baseArrayLayer == other.baseArrayLayer &&
+                       layerCount == other.layerCount &&
+                       viewType == other.viewType &&
+                       format == other.format;
+            }
+        };
+
+        struct SampledImageViewKey {
+            Uint32 baseMipLevel = 0;
+            Uint32 levelCount = 1;
+            VkImageViewType viewType = VK_IMAGE_VIEW_TYPE_2D;
+            VkFormat format = VK_FORMAT_UNDEFINED;
+
+            Bool operator==(const SampledImageViewKey& other) const {
+                return baseMipLevel == other.baseMipLevel &&
+                       levelCount == other.levelCount &&
+                       viewType == other.viewType &&
+                       format == other.format;
+            }
+        };
+
+        struct SampledImageViewKeyHash {
+            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>{}(static_cast<Uint32>(key.viewType)) +
+                        0x9e3779b9u + (hash << 6) + (hash >> 2);
+                hash ^= std::hash<Uint32>{}(static_cast<Uint32>(key.format)) +
+                        0x9e3779b9u + (hash << 6) + (hash >> 2);
+                return hash;
+            }
+        };
+
+        struct StorageImageViewKeyHash {
+            SizeT operator()(const StorageImageViewKey& key) const {
+                SizeT hash = std::hash<Uint32>{}(key.mipLevel);
+                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);
+                return hash;
+            }
+        };
+
         VkImage image = VK_NULL_HANDLE;
         VmaAllocation allocation = nullptr;
         VkImageView fullView = VK_NULL_HANDLE;
+        VkImageView sampledView = VK_NULL_HANDLE;
         Vector<VkImageView> perMipViews;
+        Vector<VkImageView> perMipSampledViews;
+        UnorderedMap<AttachmentViewKey, VkImageView, AttachmentViewKeyHash> attachmentViews;
+        UnorderedMap<SampledImageViewKey, VkImageView, SampledImageViewKeyHash> alternateSampledViews;
+        UnorderedMap<StorageImageViewKey, VkImageView, StorageImageViewKeyHash> storageImageViews;
         VkImageLayout layout = VK_IMAGE_LAYOUT_UNDEFINED;
         VkExtent2D extent = {0, 0};
+        Uint32 depth = 1;
+        Uint32 arrayLayers = 1;
         Uint32 mipLevels = 1;
         Uint32 sampledBaseMipLevel = 0;
         Uint32 sampledLevelCount = 1;
         VkFormat format = VK_FORMAT_UNDEFINED;
         VkImageAspectFlags aspect = VK_IMAGE_ASPECT_NONE;
+        VkImageViewType viewType = VK_IMAGE_VIEW_TYPE_2D;
+        VkSampleCountFlagBits sampleCount = VK_SAMPLE_COUNT_1_BIT;
+        VkImageCreateFlags imageCreateFlags = 0;
         Uint16 syncedTextureParamsVersion = 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;
+        // Snapshot of the defined mip-level count at the last sync. Folded into the early-out key
+        // 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;
 
         TextureResource() = default;
         TextureResource(const TextureResource&) = delete;
@@ -48,41 +172,88 @@ public:
             std::swap(this->image, that.image);
             std::swap(this->allocation, that.allocation);
             std::swap(this->fullView, that.fullView);
+            std::swap(this->sampledView, that.sampledView);
             std::swap(this->perMipViews, that.perMipViews);
+            std::swap(this->perMipSampledViews, that.perMipSampledViews);
+            std::swap(this->attachmentViews, that.attachmentViews);
+            std::swap(this->alternateSampledViews, that.alternateSampledViews);
+            std::swap(this->storageImageViews, that.storageImageViews);
             std::swap(this->layout, that.layout);
             std::swap(this->extent, that.extent);
+            std::swap(this->depth, that.depth);
+            std::swap(this->arrayLayers, that.arrayLayers);
             std::swap(this->mipLevels, that.mipLevels);
             std::swap(this->sampledBaseMipLevel, that.sampledBaseMipLevel);
             std::swap(this->sampledLevelCount, that.sampledLevelCount);
             std::swap(this->format, that.format);
             std::swap(this->aspect, that.aspect);
+            std::swap(this->viewType, that.viewType);
+            std::swap(this->sampleCount, that.sampleCount);
+            std::swap(this->imageCreateFlags, that.imageCreateFlags);
             std::swap(this->syncedTextureParamsVersion, that.syncedTextureParamsVersion);
+            std::swap(this->syncedContentVersion, that.syncedContentVersion);
+            std::swap(this->syncedMipLevelCount, that.syncedMipLevelCount);
         }
 
         void Reset() {
             if (fullView != VK_NULL_HANDLE) {
                 vkDestroyImageView(s_device, fullView, nullptr);
             }
+            if (sampledView != VK_NULL_HANDLE) {
+                vkDestroyImageView(s_device, sampledView, nullptr);
+            }
             for (const auto attachmentView : perMipViews) {
                 if (attachmentView != VK_NULL_HANDLE) {
                     vkDestroyImageView(s_device, attachmentView, nullptr);
                 }
             }
+            for (const auto sampledView : perMipSampledViews) {
+                if (sampledView != VK_NULL_HANDLE) {
+                    vkDestroyImageView(s_device, sampledView, nullptr);
+                }
+            }
+            for (const auto& [_, attachmentView] : attachmentViews) {
+                if (attachmentView != VK_NULL_HANDLE) {
+                    vkDestroyImageView(s_device, attachmentView, nullptr);
+                }
+            }
+            for (const auto& [_, sampledView] : alternateSampledViews) {
+                if (sampledView != VK_NULL_HANDLE) {
+                    vkDestroyImageView(s_device, sampledView, nullptr);
+                }
+            }
+            for (const auto& [_, storageImageView] : storageImageViews) {
+                if (storageImageView != VK_NULL_HANDLE) {
+                    vkDestroyImageView(s_device, storageImageView, nullptr);
+                }
+            }
             if (image != VK_NULL_HANDLE && allocation != nullptr) {
                 vmaDestroyImage(s_allocator, image, allocation);
             }
             fullView = VK_NULL_HANDLE;
+            sampledView = VK_NULL_HANDLE;
             perMipViews.clear();
+            perMipSampledViews.clear();
+            attachmentViews.clear();
+            alternateSampledViews.clear();
+            storageImageViews.clear();
             image = VK_NULL_HANDLE;
             allocation = nullptr;
             layout = VK_IMAGE_LAYOUT_UNDEFINED;
             extent = {0, 0};
+            depth = 1;
+            arrayLayers = 1;
             mipLevels = 1;
             sampledBaseMipLevel = 0;
             sampledLevelCount = 1;
             format = VK_FORMAT_UNDEFINED;
             aspect = VK_IMAGE_ASPECT_NONE;
+            viewType = VK_IMAGE_VIEW_TYPE_2D;
+            sampleCount = VK_SAMPLE_COUNT_1_BIT;
+            imageCreateFlags = 0;
             syncedTextureParamsVersion = 0;
+            syncedContentVersion = 0;
+            syncedMipLevelCount = 0;
         }
 
         ~TextureResource() {
@@ -95,21 +266,71 @@ public:
 
     Bool Initialize(const InitInfo& initInfo);
     void Shutdown();
+    void BeginFrame(Uint32 frameIndex);
 
     TextureResource* SyncTextureAndGetDescriptor(
         MG_State::GLState::ITextureObject& texture);
     VkImageView GetOrCreateViewAtMipLevel(MG_State::GLState::ITextureObject& texture, Uint32 mipLevel);
+    VkImageView GetOrCreateAttachmentViewAtMipLevel(MG_State::GLState::ITextureObject& texture, Uint32 mipLevel,
+                                                    Uint32 baseArrayLayer, Uint32 layerCount,
+                                                    VkImageViewType viewType);
+    VkImageView GetOrCreateSampledViewAtMipLevel(MG_State::GLState::ITextureObject& texture, Uint32 mipLevel);
+    VkImageView GetOrCreateSampledImageView(MG_State::GLState::ITextureObject& texture, VkFormat format);
+    VkImageView GetOrCreateStorageImageView(MG_State::GLState::ITextureObject& texture, Uint32 mipLevel,
+                                            VkFormat format, Bool layered, Int32 layer);
     void UpdateTrackedImageLayout(MG_State::GLState::ITextureObject* texture, VkImageLayout newLayout);
+    void UpdateTrackedImageLayoutAfterAttachmentWrite(VkCommandBuffer commandBuffer,
+                                                      MG_State::GLState::ITextureObject* texture,
+                                                      Uint32 writtenMipLevel,
+                                                      VkImageLayout newLayout);
     Bool TransitionTextureForSampling(VkCommandBuffer commandBuffer, MG_State::GLState::ITextureObject& texture);
+    Bool TransitionTextureForStorageImage(VkCommandBuffer commandBuffer, MG_State::GLState::ITextureObject& texture);
+    // 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;
+
+    static VkImageAspectFlags ResolveSampledImageViewAspectMask(VkImageAspectFlags imageAspect);
+    static VkFormat ResolveSampledImageViewFormat(VkFormat imageFormat, SamplerNumericDomain numericDomain);
+    static Bool AreSampledImageViewFormatsCompatible(VkFormat imageFormat, VkFormat viewFormat);
+    static Bool AreStorageImageViewFormatsCompatible(VkFormat imageFormat, VkFormat viewFormat);
 
     static Bool TransitionImageLayout(VkCommandBuffer commandBuffer, VkImage image, VkImageLayout& trackedLayout,
                                VkImageLayout newLayout, VkPipelineStageFlags srcStageMask,
                                VkPipelineStageFlags dstStageMask, VkAccessFlags srcAccessMask,
                                VkAccessFlags dstAccessMask, VkImageAspectFlags aspectMask,
-                               Uint32 baseMipLevel = 0, Uint32 levelCount = 1);
+                               Uint32 baseMipLevel = 0, Uint32 levelCount = 1,
+                               Uint32 layerCount = 1);
 
     SizeT CollectGarbage();
+
+    // Per-draw sync memo. Within a single SetupDraw the same sampled texture is
+    // resolved ~3x (SetupDraw's layout-probe loop, its post-transition loop, and
+    // again inside ResolveSamplerDescriptor). No GL texture mutation can happen
+    // mid-SetupDraw, and layout is tracked on the TextureResource independently of
+    // SyncTexture, so after the first successful sync of a texture in a draw the
+    // heavy SyncTexture work (mip-completeness/resource/view resync + dirty scan)
+    // is pure redundancy. BeginDrawSyncScope opens a window in which repeat
+    // SyncTextureAndGetDescriptor calls short-circuit to the already-synced
+    // resource; EndDrawSyncScope closes it. Use the RAII DrawSyncScope guard.
+    void BeginDrawSyncScope();
+    void EndDrawSyncScope();
+
+    // RAII guard that opens/closes a per-draw sync memo window (see above).
+    class DrawSyncScope {
+    public:
+        explicit DrawSyncScope(VkTextureManager& manager) : m_manager(manager) { m_manager.BeginDrawSyncScope(); }
+        ~DrawSyncScope() { m_manager.EndDrawSyncScope(); }
+        DrawSyncScope(const DrawSyncScope&) = delete;
+        DrawSyncScope& operator=(const DrawSyncScope&) = delete;
+    private:
+        VkTextureManager& m_manager;
+    };
+
 private:
+    // Bumped in SyncTextureResource right after vmaCreateImage(texture). See GetTextureImageEpoch().
+    Uint64 m_textureImageEpoch = 1;
 
     Bool SyncTexture(MG_State::GLState::ITextureObject &texture,
                      TextureResource &outResource);
@@ -119,7 +340,11 @@ private:
                              TextureResource &resource);
     Bool SyncTextureViews(const MG_State::GLState::ITextureObject& texture, TextureResource& resource);
     VkImageView CreateImageView(VkImage image, VkFormat format, VkImageAspectFlags aspect,
-                                Uint32 baseMipLevel, Uint32 levelCount) const;
+                                VkImageViewType viewType, Uint32 baseMipLevel, Uint32 levelCount,
+                                Uint32 baseArrayLayer,
+                                Uint32 layerCount,
+                                const VkComponentMapping* components = nullptr,
+                                VkImageUsageFlags viewUsage = 0) const;
     Bool UploadDirtyMipLevels(MG_State::GLState::TextureObjectMipmap &mipmapTexture,
                       TextureUploadTarget uploadTarget,
                       TextureResource &outResource);
@@ -132,15 +357,40 @@ private:
     static void ResolveViewMipRange(const MG_State::GLState::ITextureObject& texture, Uint32 mipLevels,
                                     Uint32& outBaseMipLevel, Uint32& outLevelCount);
     static VkImageAspectFlags GetAspectMaskForFormat(VkFormat format);
+    void DeferResourceRelease(TextureResource&& resource);
+    void DeferViewRelease(VkImageView view);
+    void CollectDeferredReleases(Uint32 frameIndex);
+    void DestroyDeferredReleases();
+    static TextureIdentity MakeTextureIdentity(MG_State::GLState::ITextureObject* texture);
+    void EraseTrackedTexture(const TextureIdentity& identity);
+    void PruneStaleTextureAliases(MG_State::GLState::ITextureObject* texture);
 
     VkDevice m_device = VK_NULL_HANDLE;
     VkPhysicalDevice m_physicalDevice = VK_NULL_HANDLE;
     VmaAllocator m_allocator = nullptr;
     VkCommandPool m_commandPool = VK_NULL_HANDLE;
     VkQueue m_graphicsQueue = VK_NULL_HANDLE;
+    Uint32 m_currentFrameIndex = 0;
 
     Uint8 m_gcCounter = 0;
-    UnorderedMap<MG_State::GLState::ITextureObject*, WeakPtr<MG_State::GLState::ITextureObject>> m_aliveObjects;
-    UnorderedMap<MG_State::GLState::ITextureObject*, TextureResource> m_textureResources;
+    // Active only between BeginDrawSyncScope/EndDrawSyncScope; identities of
+    // textures already fully synced in the current draw (small N -> flat scan).
+    Bool m_drawSyncScopeActive = false;
+    // Per-draw sync memo: the identity plus the resolved resource pointer. The pointer is stable
+    // across rehash in the node-based m_textureResources and stays valid for the draw (a texture
+    // synced this draw is alive and is not erased mid-draw), so a repeat sync of the same texture
+    // returns the resource without re-hashing the identity into m_textureResources.
+    struct DrawSyncedTexture {
+        TextureIdentity identity;
+        TextureResource* resource = nullptr;
+    };
+    Vector<DrawSyncedTexture> m_drawSyncedThisDraw;
+    // 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;
+    std::unordered_map<TextureIdentity, WeakPtr<MG_State::GLState::ITextureObject>, TextureIdentityHash> m_aliveObjects;
+    std::unordered_map<TextureIdentity, TextureResource, TextureIdentityHash> m_textureResources;
+    Vector<Vector<TextureResource>> m_deferredReleases;
+    Vector<Vector<VkImageView>> m_deferredViewReleases;
 };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkTimerQueryManager.cpp

@@ -0,0 +1,179 @@
+// MobileGL - MobileGL/MG_Backend/DirectVulkan/Renderer/VkTimerQueryManager.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 "VkTimerQueryManager.h"
+
+namespace MobileGL::MG_Backend::DirectVulkan {
+    Bool VkTimerQueryManager::Initialize(const InitInfo& initInfo) {
+        Shutdown();
+
+        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)",
+                    initInfo.timestampValidBits, initInfo.timestampPeriodNs, initInfo.slotsPerPool);
+            return false;
+        }
+
+        m_device = initInfo.device;
+        m_timestampPeriodNs = initInfo.timestampPeriodNs;
+        m_validBitsMask = initInfo.timestampValidBits >= 64
+                              ? ~0ull
+                              : ((1ull << initInfo.timestampValidBits) - 1ull);
+        m_slotsPerPool = initInfo.slotsPerPool;
+        m_pools.resize(initInfo.frameCount);
+
+        VkQueryPoolCreateInfo poolInfo{};
+        poolInfo.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO;
+        poolInfo.queryType = VK_QUERY_TYPE_TIMESTAMP;
+        poolInfo.queryCount = m_slotsPerPool;
+        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));
+                Shutdown();
+                return false;
+            }
+        }
+        return true;
+    }
+
+    void VkTimerQueryManager::Shutdown() {
+        if (m_device != VK_NULL_HANDLE) {
+            for (auto& poolState : m_pools) {
+                if (poolState.pool != VK_NULL_HANDLE) {
+                    vkDestroyQueryPool(m_device, poolState.pool, nullptr);
+                }
+            }
+        }
+        // Records the frontend still holds simply stay unharvested; their
+        // results read back as 0.
+        m_pools.clear();
+        m_device = VK_NULL_HANDLE;
+        m_timestampPeriodNs = 0.0f;
+        m_validBitsMask = 0;
+        m_slotsPerPool = 0;
+    }
+
+    void VkTimerQueryManager::OnFrameCommandRecordingBegan(VkCommandBuffer commandBuffer, Uint32 frameIndex,
+                                                           Uint64 frameSerial) {
+        MOBILEGL_ASSERT(frameIndex < m_pools.size(), "VkTimerQueryManager frame index out of range");
+        auto& poolState = m_pools[frameIndex];
+        if (poolState.preparedFrameSerial == frameSerial) {
+            // Recording re-began within the same frame (mid-frame readback
+            // submit or the Present layout transition); the pool was already
+            // harvested and reset for this cycle, and resetting again would
+            // clobber timestamps written earlier in the frame.
+            return;
+        }
+
+        // Harvest what the pool's previous cycle left behind. The frame slot's
+        // fence was waited before re-recording, so every executed query is
+        // already available and the reads return immediately.
+        DrainPoolPending(poolState);
+
+        vkCmdResetQueryPool(commandBuffer, poolState.pool, 0, m_slotsPerPool);
+        poolState.cursor = 0;
+        poolState.exhaustionWarned = false;
+        poolState.preparedFrameSerial = frameSerial;
+    }
+
+    SharedPtr<VkTimerQueryManager::TimestampRecord> VkTimerQueryManager::WriteTimestamp(VkCommandBuffer commandBuffer,
+                                                                                        Uint32 frameIndex,
+                                                                                        Uint64 frameSerial) {
+        MOBILEGL_ASSERT(frameIndex < m_pools.size(), "VkTimerQueryManager frame index out of range");
+        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 "
+                        "this frame fall back to the frontend path",
+                        frameIndex, m_slotsPerPool);
+                poolState.exhaustionWarned = true;
+            }
+            return nullptr;
+        }
+
+        auto record = MakeShared<TimestampRecord>();
+        record->poolIndex = frameIndex;
+        record->slot = poolState.cursor++;
+        record->frameSerial = frameSerial;
+        vkCmdWriteTimestamp(commandBuffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, poolState.pool, record->slot);
+        poolState.pendingRecords.push_back(record);
+        return record;
+    }
+
+    Bool VkTimerQueryManager::TryHarvest(TimestampRecord& record) {
+        if (record.harvested) {
+            return true;
+        }
+        if (m_device == VK_NULL_HANDLE || record.poolIndex >= m_pools.size()) {
+            return false;
+        }
+
+        Uint64 resultWithAvailability[2] = {0, 0};
+        const VkResult result = vkGetQueryPoolResults(
+            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));
+            return false;
+        }
+        if (resultWithAvailability[1] == 0) {
+            return false;
+        }
+        record.rawTicks = resultWithAvailability[0];
+        record.harvested = true;
+        return true;
+    }
+
+    void VkTimerQueryManager::InvalidatePendingRecords() {
+        for (auto& poolState : m_pools) {
+            DrainPoolPending(poolState);
+            // Force a harvest-free reset cycle the next time this pool's frame
+            // begins recording.
+            poolState.preparedFrameSerial = 0;
+        }
+    }
+
+    void VkTimerQueryManager::DrainPoolPending(PoolState& poolState) {
+        for (auto& record : poolState.pendingRecords) {
+            if (record->harvested) {
+                continue;
+            }
+            if (!TryHarvest(*record)) {
+                // The commands carrying this timestamp never executed (they
+                // were dropped, e.g. by a swapchain recreation mid-frame).
+                // Mark the record resolved-as-invalid so waits on it cannot
+                // hang; its result reads back as 0.
+                record->harvested = true;
+                record->valid = false;
+            }
+        }
+        poolState.pendingRecords.clear();
+    }
+
+    Uint64 VkTimerQueryManager::MaskToValidBits(Uint64 ticks) const {
+        return ticks & m_validBitsMask;
+    }
+
+    Uint64 VkTimerQueryManager::ElapsedNs(const TimestampRecord& begin, const TimestampRecord& end) const {
+        if (!begin.valid || !end.valid) {
+            return 0;
+        }
+        const Uint64 deltaTicks = MaskToValidBits(end.rawTicks - begin.rawTicks);
+        return static_cast<Uint64>(static_cast<double>(deltaTicks) * static_cast<double>(m_timestampPeriodNs));
+    }
+
+    Uint64 VkTimerQueryManager::TimestampNs(const TimestampRecord& record) const {
+        if (!record.valid) {
+            return 0;
+        }
+        return static_cast<Uint64>(static_cast<double>(MaskToValidBits(record.rawTicks)) *
+                                   static_cast<double>(m_timestampPeriodNs));
+    }
+} // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VkTimerQueryManager.h

@@ -0,0 +1,111 @@
+// MobileGL - MobileGL/MG_Backend/DirectVulkan/Renderer/VkTimerQueryManager.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 "../VkIncludes.h"
+#include <Includes.h>
+
+namespace MobileGL::MG_Backend::DirectVulkan {
+    // GPU timestamp storage backing the GL timer-query frontend (GL_TIME_ELAPSED
+    // spans and GL_TIMESTAMP one-shots): one VkQueryPool of timestamp slots per
+    // frame in flight.
+    //
+    // Per-frame lifecycle: right after a frame slot's command buffer begins
+    // recording (and before any render pass, since vkCmdResetQueryPool must be
+    // recorded outside one), OnFrameCommandRecordingBegan harvests every
+    // not-yet-read slot of the pool about to be reused (the slot's frame fence
+    // was waited before re-recording, so the results are already available),
+    // records a reset of the whole pool, and rewinds the allocation cursor.
+    class VkTimerQueryManager {
+    public:
+        // One vkCmdWriteTimestamp landing spot. Shared (via SharedPtr) between
+        // the frontend-held query object and the owning pool's pending list, so
+        // deleting a query while its result is still in flight never leaves the
+        // pool with a dangling record.
+        struct TimestampRecord {
+            Uint32 poolIndex = 0;
+            Uint32 slot = 0;
+            // VkBufferManager frame serial current when the timestamp was
+            // recorded; result availability is bounded by its completion.
+            Uint64 frameSerial = 0;
+            Bool harvested = false;
+            // Cleared when the recorded commands were dropped before they could
+            // execute (swapchain recreation abandons the in-progress command
+            // buffer); the result then reads back as 0.
+            Bool valid = true;
+            Uint64 rawTicks = 0;
+        };
+
+        struct InitInfo {
+            VkDevice device = VK_NULL_HANDLE;
+            Uint32 frameCount = 0;
+            Uint32 timestampValidBits = 0;
+            Float timestampPeriodNs = 0.0f; // nanoseconds per timestamp tick
+            Uint32 slotsPerPool = 128;
+        };
+
+        Bool Initialize(const InitInfo& initInfo);
+        // The caller guarantees the device is idle (same contract as the other
+        // DirectVulkan managers' Shutdown paths).
+        void Shutdown();
+
+        // The per-frame hook described in the class comment. Re-begins within
+        // the same frame serial (mid-frame readback submits, the Present layout
+        // transition) are skipped so already-written slots survive.
+        void OnFrameCommandRecordingBegan(VkCommandBuffer commandBuffer, Uint32 frameIndex, Uint64 frameSerial);
+
+        // Allocates a slot from the frame's pool and records a bottom-of-pipe
+        // vkCmdWriteTimestamp (valid both inside and outside a render pass).
+        // Returns null on pool exhaustion, with one warning per pool cycle; the
+        // frontend falls back gracefully on a null handle.
+        SharedPtr<TimestampRecord> WriteTimestamp(VkCommandBuffer commandBuffer, Uint32 frameIndex,
+                                                  Uint64 frameSerial);
+
+        // Non-blocking single-slot read (WITH_AVAILABILITY, no WAIT). Returns
+        // true once the record holds its raw ticks. Callers gate this on the
+        // record's frame serial being complete.
+        Bool TryHarvest(TimestampRecord& record);
+
+        // Reads every pending result that is available (the caller guarantees
+        // the device is idle) and marks the rest invalid. Called when recorded
+        // but unsubmitted commands are dropped (swapchain recreation), which
+        // would otherwise leave slots that never become available. Each pool is
+        // reset lazily on its next OnFrameCommandRecordingBegan.
+        void InvalidatePendingRecords();
+
+        // end - begin using unsigned wrap arithmetic masked to the queue's
+        // timestampValidBits, converted to nanoseconds. 0 if either record was
+        // invalidated.
+        Uint64 ElapsedNs(const TimestampRecord& begin, const TimestampRecord& end) const;
+        // Raw GPU timestamp converted to nanoseconds. 0 if invalidated.
+        Uint64 TimestampNs(const TimestampRecord& record) const;
+
+    private:
+        struct PoolState {
+            VkQueryPool pool = VK_NULL_HANDLE;
+            Uint32 cursor = 0;
+            // Frame serial the pool was last harvested + reset for; guards
+            // against double resets when recording re-begins mid-frame.
+            Uint64 preparedFrameSerial = 0;
+            Bool exhaustionWarned = false;
+            Vector<SharedPtr<TimestampRecord>> pendingRecords;
+        };
+
+        Uint64 MaskToValidBits(Uint64 ticks) const;
+        // Harvest (or invalidate, when the result never became available)
+        // every pending record of a pool and clear its pending list.
+        void DrainPoolPending(PoolState& pool);
+
+        VkDevice m_device = VK_NULL_HANDLE;
+        Float m_timestampPeriodNs = 0.0f;
+        Uint64 m_validBitsMask = 0;
+        Uint32 m_slotsPerPool = 0;
+        Vector<PoolState> m_pools;
+    };
+} // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.h

@@ -20,6 +20,7 @@
 #include "VkRenderPassManager.h"
 #include "VkSamplerManager.h"
 #include "VkTextureManager.h"
+#include "VkTimerQueryManager.h"
 #include "MG_Util/Math/VectorTypes.h"
 #include <Includes.h>
 #include <vk_mem_alloc.h>
@@ -50,6 +51,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         Uint32 instanceCount = 1;
         Uint32 firstVertex = 0;
         Uint32 firstInstance = 0;
+        // Indexed-draw metadata for bounding vertex-stream conversion. baseVertex is the
+        // draw's base-vertex offset; indexRangeIsExactView is true only when the draw
+        // fetches exactly the indices its IndexBufferView describes (direct DrawElements;
+        // multi/indirect forms leave it false because the CPU cannot bound their ranges).
+        Int32 baseVertex = 0;
+        Bool indexRangeIsExactView = false;
     };
 
     struct DrawIndexedCmdParam {
@@ -86,6 +93,12 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         DrawIndexedCmdParam* pParams = nullptr;
     };
 
+    struct MultiDrawCmd {
+        GLenum mode = GL_TRIANGLES;
+        Uint32 drawCount = 0;
+        DrawCmdParam* pParams = nullptr;
+    };
+
     struct QueueFamilyIndices {
         Int32 graphicsFamily = -1;
         Int32 presentFamily = -1;
@@ -101,7 +114,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         }
     };
 
-    class VulkanRenderer {
+    class VulkanRenderer : public IBufferCopyCommandProvider, public FrameContext::IRecordingObserver {
     public:
         VulkanRenderer(NativeWindowType window, const VulkanRendererConfig& cfg = {});
         ~VulkanRenderer();
@@ -109,24 +122,144 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         void Initialize();
         void Shutdown();
 
+        // IBufferCopyCommandProvider: recording command buffer, outside any
+        // render pass, for immediate staged buffer copies.
+        VkCommandBuffer AcquireBufferCopyCommandBuffer() override;
+
+        // FrameContext::IRecordingObserver: prepares the frame's timer-query
+        // pool (harvest + reset) right after the frame command buffer begins
+        // recording, before any render pass.
+        void OnFrameCommandRecordingBegan(VkCommandBuffer commandBuffer) override;
+
         Bool SetupDraw(FrameContext::FrameData& frame, GLenum mode, Flags<DrawSetupAspect> aspects,
+                       const DrawCmdParam& drawParams,
                        const IndexBufferView* pIndexBufferView = nullptr);
         void ClearAttachmentsOnActiveRenderPass(VkCommandBuffer commandBuffer,
                                                 const RenderPassEntry& compatibleRenderPassEntry);
 
+        enum class ScissoredClearPrep {
+            NotNeeded,  // scissor covers the whole target — take the deferred whole-surface path instead
+            NoOp,       // nothing to clear (degenerate target or empty scissor rect)
+            Ready,      // a render pass is active; record vkCmdClearAttachments with the returned rect
+        };
+        ScissoredClearPrep PrepareScissoredClear(const MG_State::GLState::FramebufferObject& framebuffer,
+                                                 VkClearRect& outClearRect);
+
         void Clear(GLbitfield mask);
+        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 ClearNamedFramebufferfi(const SharedPtr<MG_State::GLState::FramebufferObject>& framebuffer,
+                                     GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
         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>& readFbo,
+                                  const SharedPtr<MG_State::GLState::FramebufferObject>& drawFbo,
+                                  GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                                  GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                                  GLbitfield mask, GLenum filter);
+        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,
+                              GLenum srcTarget, GLint srcLevel, GLint srcX, GLint srcY, GLint srcZ,
+                              const SharedPtr<MG_State::GLState::ITextureObject>& dstTexture,
+                              GLenum dstTarget, GLint dstLevel, GLint dstX, GLint dstY, GLint dstZ,
+                              GLsizei srcWidth, GLsizei srcHeight, GLsizei srcDepth);
+        void GenerateMipmap(GLenum target);
+        void ReadPixels(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, void* pixels);
+        static SizeT GetReadbackTexelSize(VkFormat sourceFormat);
+        static Bool ConvertReadbackPixels(const Uint8* sourcePixels, VkFormat sourceFormat,
+                                          GLsizei width, GLsizei height, GLenum destinationFormat,
+                                          GLenum destinationType, SizeT destinationRowStride,
+                                          Uint8* destinationPixels);
+        void GetTexImage(GLenum target, GLint level, GLenum format, GLenum type, GLvoid* pixels);
+        void GetTextureImage(const SharedPtr<MG_State::GLState::ITextureObject>& texture,
+                             TextureUploadTarget uploadTarget, GLint level, GLenum format, GLenum type,
+                             GLsizei bufSize, GLvoid* pixels);
+        void DispatchCompute(GLuint numGroupsX, GLuint numGroupsY, GLuint numGroupsZ);
+        void DispatchComputeIndirect(GLintptr indirect);
+        void MemoryBarrier(GLbitfield barriers);
+        static VkMemoryBarrier BuildMemoryBarrierForGlBarriers(GLbitfield barriers);
         void DrawArrays(const DrawCmd& payload);
         void DrawElements(const DrawIndexedCmd& payload);
+        void MultiDrawArrays(const MultiDrawCmd& payload);
         void MultiDrawElements(const MultiDrawIndexedCmd& payloads);
+        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,
+                                            GLsizei maxdrawcount, GLsizei stride);
         void Present();
 
         const PhysicalDevice& GetPhysicalDevice() const;
+        VkInstance GetInstance() const;
         Bool IsDrawIndirectCountExtensionEnabled() const;
 
-        void RecreateSwapchain();
+        // GL fence support, expressed in queue-submission indices backed by
+        // real VkFences. A GL fence captures GetSyncPointSubmitIndex() at
+        // creation: the index of the submission that will carry the commands
+        // recorded so far (m_submitCounter + 1 while work is pending, or
+        // m_submitCounter when nothing has been recorded since the last
+        // submit). It is signaled once that submission's fence is observed
+        // signaled - unlike the frame-serial heuristic, this makes fences
+        // signal as soon as the GPU actually finishes, which MC 1.21.5's
+        // fence-paced ring buffers rely on to recycle their space.
+        Uint64 GetSyncPointSubmitIndex() const;
+        // Non-blocking: polls outstanding submission fences and reports
+        // whether every submission up to `submitIndex` has completed.
+        Bool IsSubmitIndexComplete(Uint64 submitIndex);
+        // Submits the commands recorded so far without waiting (GL flush).
+        // Recording restarts lazily on a fresh command buffer; the submitted
+        // one is retired until the frame slot's fence is next waited. Returns
+        // true when a submission was made.
+        Bool FlushPendingCommands();
+        // Flush gated on usefulness: only flushes when `submitIndex` is still
+        // unsubmitted, so poll loops on already-submitted fences do not split
+        // the frame's render pass (a full tile load/store on TBDR GPUs).
+        Bool FlushForSyncPoint(Uint64 submitIndex);
+        // Blocking wait for a submission index with a nanosecond timeout.
+        // When the index is still unsubmitted and flushIfPending is set, the
+        // pending commands are flushed first so the wait can make progress.
+        Bool WaitForSubmitIndex(Uint64 submitIndex, Uint64 timeoutNs, Bool flushIfPending);
+
+        // Frame-serial completion, still used by the timer-query paths (their
+        // records are bucketed per frame slot).
+        Bool IsFrameSerialComplete(Uint64 serial) const;
+        // Blocking wait for a submitted serial. Returns false when the serial
+        // cannot complete without further submissions (it belongs to the
+        // current, not-yet-presented frame) or when the wait failed.
+        Bool WaitForFrameSerial(Uint64 serial, Uint64 timeoutNs);
+
+        // GPU timer queries, backing the GL_TIME_ELAPSED / GL_TIMESTAMP
+        // frontend. Timestamp support (queue timestampValidBits > 0 and a
+        // non-zero timestampPeriod) is cached at device creation.
+        Bool IsTimerQuerySupported() const;
+        // The samplerAnisotropy device feature was granted, so GL_TEXTURE_MAX_ANISOTROPY_EXT is
+        // honored rather than accepted-and-ignored.
+        Bool IsSamplerAnisotropySupported() const { return m_samplerAnisotropyFeatureEnabled; }
+        // Ensures the frame command buffer is recording (same lazy pattern as
+        // SetupDraw) and writes a bottom-of-pipe timestamp into the current
+        // frame's pool. Null when unsupported or the pool is exhausted.
+        SharedPtr<VkTimerQueryManager::TimestampRecord> WriteTimerQueryTimestamp();
+        // Non-blocking: true once the record's raw ticks are on the CPU
+        // (harvests the slot once its frame serial has completed).
+        Bool IsTimerQueryResultReady(VkTimerQueryManager::TimestampRecord& record);
+        // Blocking wait, mirroring ClientWaitSync's caveat: a record written
+        // this frame cannot complete until Present submits the commands, so
+        // this returns false (result reads as 0) instead of deadlocking.
+        Bool WaitForTimerQueryResult(VkTimerQueryManager::TimestampRecord& record);
+        Uint64 GetTimerQueryElapsedNs(const VkTimerQueryManager::TimestampRecord& begin,
+                                      const VkTimerQueryManager::TimestampRecord& end) const;
+        Uint64 GetTimerQueryTimestampNs(const VkTimerQueryManager::TimestampRecord& record) const;
+
+        void RequestSwapchainResize(Uint32 width, Uint32 height);
+        // Returns false when the surface is zero-area (minimized/hidden window):
+        // no new swapchain is installed and presentation must stay suspended.
+        Bool RecreateSwapchain();
 
     private:
         struct BlitUniformData {
@@ -146,8 +279,88 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             Uint32 samplerBinding = 0;
         };
 
+        struct DepthMipmapResources {
+            SharedPtr<MG_State::GLState::ProgramObject> program;
+            Int srcRectLocation = -1;
+            Int dstRectLocation = -1;
+            Int surfaceTransformLocation = -1;
+            Int srcTexelSizeLocation = -1;
+            Uint32 samplerBinding = 0;
+        };
+
+        struct DeferredDepthMipmapCleanup {
+            Vector<VkImageView> imageViews;
+            Vector<VkFramebuffer> framebuffers;
+            Vector<VkRenderPass> renderPasses;
+            Vector<VkPipeline> pipelines;
+        };
+
+        void QueueClearBufferPayload(GLenum buffer, GLint drawbuffer, const ClearAttachmentPayload& clearPayload);
+        void QueueClearBufferPayloadForFramebuffer(const MG_State::GLState::FramebufferObject& framebuffer,
+                                                  GLenum buffer, GLint drawbuffer,
+                                                  const ClearAttachmentPayload& clearPayload);
+        void RecordScissoredClearBuffer(const MG_State::GLState::FramebufferObject& framebuffer,
+                                        GLenum buffer, GLint drawbuffer,
+                                        const ClearAttachmentPayload& clearPayload,
+                                        const VkClearRect& clearRect);
+
+        // ---- Submission fence tracking (GL sync objects) ----
+        // One record per vkQueueSubmit still in flight, in ascending submit
+        // order. Present/readback submissions reference the frame slot's
+        // fence (not pool-owned); mid-frame flushes use pooled fences that are
+        // recycled once their submission is observed complete.
+        // Not thread-safe: like the rest of the renderer, the tracker relies
+        // on GL calls being serialized (launchers migrate the context across
+        // threads, but calls never run concurrently), so sync-object polls
+        // may mutate it without locking.
+        struct SubmitRecord {
+            Uint64 submitIndex = 0;
+            // Buffer-manager frame serial the submission was made under; its
+            // completion raises the completed-serial floor (timer queries and
+            // buffer busy-tracking live in frame-serial space).
+            Uint64 frameSerial = 0;
+            VkFence fence = VK_NULL_HANDLE;
+            Bool pooledFence = false;
+        };
+        // Registers a submission that vkQueueSubmit just made with `fence`.
+        // Invariant: every graphics-queue submission that outlives its call
+        // site must be registered so GL fences observe it. Exempt are the
+        // texture-upload/preserve submits in VkTextureManager, which
+        // vkWaitForFences inline before returning.
+        void RegisterSubmit(VkFence fence, Bool pooledFence);
+        // Builds the submit packet for the frame's pending command buffer
+        // (consuming the acquire semaphore on the slot's first submission),
+        // submits it with `fence`, and registers the submission. On failure
+        // the frame state is left untouched. Shared by the mid-frame flush
+        // and the readback path so the semaphore-consumption invariant lives
+        // in one place.
+        Bool SubmitPendingCommandBuffer(FrameContext::FrameData& frame, VkFence fence, Bool pooledFence);
+        // Polls in-flight submission fences (prefix order) and advances the
+        // completed counter past every fence observed signaled.
+        void RefreshCompletedSubmits();
+        // All submissions up to `submitIndex` are known complete (their fence
+        // was waited or the device was idled); drops their records and
+        // recycles pooled fences.
+        void OnSubmitsCompletedUpTo(Uint64 submitIndex);
+        VkFence AcquirePooledSubmitFence();
+        void DestroySubmitFencePool();
+        Bool HasPendingRecordedWork() const;
+
+        Vector<SubmitRecord> m_inFlightSubmits;
+        Vector<VkFence> m_freeSubmitFences;
+        Uint64 m_submitCounter = 0;
+        Uint64 m_completedSubmitCounter = 0;
+
         NativeWindowType m_window = 0;
+        void* m_platformDisplay = nullptr;
+        void* m_platformLibrary = nullptr;
+        void* m_platformCloseDisplay = nullptr;
         VulkanRendererConfig m_config;
+        Bool m_swapchainResizeRequested = false;
+        // Presentation is suspended while the window is zero-area (minimized): the
+        // swapchain is unusable/out of date, so Present drops frames instead of
+        // submitting on a signaled fence / presenting never-acquired images.
+        Bool m_presentSuspended = false;
 
         // Vulkan objects
         Bool m_validationLayersEnabled = false;
@@ -164,6 +377,27 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkQueue m_presentQueue = VK_NULL_HANDLE;
         Bool m_drawIndirectCountExtensionEnabled = false;
         Bool m_indexTypeUint8ExtensionEnabled = false;
+        Bool m_logicOpFeatureEnabled = false;
+        Bool m_multiDrawIndirectFeatureEnabled = false;
+        Bool m_samplerAnisotropyFeatureEnabled = false;
+        Bool m_shaderDrawParametersExtensionEnabled = false;
+        Bool m_shaderDrawParametersFeatureEnabled = false;
+        Bool m_unformattedFloatStorageImagesEnabled = false;
+        // fillModeNonSolid gates VK_POLYGON_MODE_LINE/_POINT (glPolygonMode); independentBlend gates
+        // per-draw-buffer color write masks (glColorMaski). Both are cached at device creation and
+        // drive a runtime fallback when the device lacks them.
+        Bool m_fillModeNonSolidFeatureEnabled = false;
+        Bool m_independentBlendFeatureEnabled = false;
+        // dualSrcBlend gates GL_SRC1_* blend factors (glBindFragDataLocationIndexed dual-source blend);
+        // primitiveTopologyListRestart gates primitive restart on *list* topologies (strip/fan restart
+        // needs no feature). Both cached at device creation and drive a hard-fail-at-draw when absent.
+        Bool m_dualSrcBlendFeatureEnabled = false;
+        Bool m_primitiveTopologyListRestartFeatureEnabled = false;
+        // Cached at device creation from the graphics queue family properties
+        // and device limits; drives timer-query support.
+        Uint32 m_timestampValidBits = 0;
+        Float m_timestampPeriodNs = 0.0f;
+        Bool m_timerQuerySupported = false;
         using PFNDrawIndexedIndirectCountFunc = void(VKAPI_PTR*)(VkCommandBuffer commandBuffer, VkBuffer buffer,
                                                                  VkDeviceSize offset, VkBuffer countBuffer,
                                                                  VkDeviceSize countBufferOffset, Uint32 maxDrawCount,
@@ -173,12 +407,24 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkCommandPool m_commandPool = VK_NULL_HANDLE;
 
         VkBufferManager m_bufferManager;
-        Vector<const MG_State::GLState::BufferObject*> m_transientVertexIndexBuffersThisFrame;
 
         Uint m_imageIndexAcquired = 0;
         FrameContext m_frameContext;
 
         UniquePtr<PipelineFactory> m_pipelineFactory;
+        // Single-slot "last pipeline" memo: skip the per-draw GetOrCreatePipeline work (state
+        // gather + synthetic vertex-input rebuild + payload hash + lookup) when the full pipeline
+        // state is unchanged from the previous draw. The key provably covers every pipeline field.
+        // Reset per-frame and on pipeline destruction so the cached handle can never dangle.
+        Bool m_lastPipelineValid = false;
+        GLenum m_lastPipelineMode = 0;
+        Uint64 m_lastPipelineProgramHash = 0;
+        Uint64 m_lastPipelineVertexInputHash = 0;
+        Uint64 m_lastPipelineRenderPassHash = 0;
+        Uint m_lastPipelineRenderStateVersion = 0;
+        ProgramFactory::CompileOptionFlags m_lastPipelineTransformFlags = {};
+        VkPipeline m_lastPipelineResult = VK_NULL_HANDLE;
+        UnorderedMap<ProgramFactory::HashType, VkPipeline> m_computePipelines;
         UniquePtr<ProgramFactory> m_programFactory;
         UniquePtr<UniformManager> m_uniformManager;
         UniquePtr<VertexInputStateFactory> m_vertexInputStateFactory;
@@ -186,7 +432,87 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         UniquePtr<VkRenderPassManager> m_renderPassManager;
         UniquePtr<VkTextureManager> m_textureManager;
         UniquePtr<VkSamplerManager> m_samplerManager;
+        UniquePtr<VkTimerQueryManager> m_timerQueryManager;
         BlitResources m_blitResources;
+        DepthMipmapResources m_depthMipmapResources;
+        Vector<DeferredDepthMipmapCleanup> m_deferredDepthMipmapCleanup;
+
+        // Skip the per-draw CollectSampledTextures walk (~5% of the render thread) when the sampled
+        // texture SET is provably unchanged from the previous draw: same program (lifetime id +
+        // backend-state version, which covers sampler-uniform reassignment / relink) and transform
+        // flags, and no texture bind/unbind/delete since (GetTextureBindGeneration). On a hit,
+        // m_sampledTexturesScratch still holds the previous draw's list and steps 2-4 (feedback /
+        // layout probe / transition) re-run on it, so layout correctness is unaffected - only the GL
+        // walk is skipped. The program lifetime id (never reused, unlike the GL name) and the
+        // monotonic bind generation make the key ABA-proof; the per-command-buffer reset is a cheap
+        // belt-and-suspenders.
+        Bool m_lastSampledSetValid = false;
+        Uint64 m_lastSampledSetProgramLifetimeId = 0;
+        Uint32 m_lastSampledSetProgramVersion = 0;
+        ProgramFactory::CompileOptionFlags m_lastSampledSetTransformFlags = {};
+        Uint64 m_lastSampledSetBindGeneration = 0;
+
+        // Per-draw scratch buffers (clear keeps capacity) — these paths run for every
+        // draw call and must not allocate.
+        Vector<MG_State::GLState::ITextureObject*> m_sampledTexturesScratch;
+        Vector<MG_State::GLState::ITextureObject*> m_storageImageTexturesScratch;
+        Vector<VkBuffer> m_vertexBuffersScratch;
+        Vector<VkDeviceSize> m_vertexOffsetsScratch;
+        Vector<VkVertexInputAttributeDescription> m_patchedAttributesScratch;
+        Vector<Float> m_vertexConversionScratch;
+        Vector<Uint8> m_vertexRepackScratch;
+
+        struct ConvertedVertexStreamKey {
+            const MG_State::GLState::BufferObject* buffer = nullptr;
+            Uint64 changeSerial = 0;
+            SizeT baseOffset = 0;
+            Uint32 sourceStride = 0;
+            DataType type = DataType::Float32;
+            Int size = 0;
+            Bool normalized = false;
+            Bool isInteger = false;
+            VertexInputStateFactory::VertexStreamConversion conversion =
+                VertexInputStateFactory::VertexStreamConversion::None;
+
+            Bool operator==(const ConvertedVertexStreamKey& other) const {
+                return buffer == other.buffer && changeSerial == other.changeSerial &&
+                       baseOffset == other.baseOffset && sourceStride == other.sourceStride &&
+                       type == other.type && size == other.size && normalized == other.normalized &&
+                       isInteger == other.isInteger && conversion == other.conversion;
+            }
+        };
+
+        struct ConvertedVertexStreamKeyHash {
+            SizeT operator()(const ConvertedVertexStreamKey& key) const {
+                SizeT hash = std::hash<const void*>{}(key.buffer);
+                auto combine = [&hash](SizeT value) {
+                    hash ^= value + static_cast<SizeT>(0x9e3779b97f4a7c15ull) + (hash << 6) + (hash >> 2);
+                };
+                combine(std::hash<Uint64>{}(key.changeSerial));
+                combine(std::hash<SizeT>{}(key.baseOffset));
+                combine(std::hash<Uint32>{}(key.sourceStride));
+                combine(std::hash<Uint32>{}(static_cast<Uint32>(key.type)));
+                combine(std::hash<Int>{}(key.size));
+                combine(std::hash<Bool>{}(key.normalized));
+                combine(std::hash<Bool>{}(key.isInteger));
+                combine(std::hash<Uint32>{}(static_cast<Uint32>(key.conversion)));
+                return hash;
+            }
+        };
+
+        struct ConvertedVertexStream {
+            BufferSlice slice;
+            // Number of source elements the cached slice covers. A draw needing a prefix of
+            // this range reuses the slice (converted streams are tightly packed); a draw
+            // needing more reconverts and replaces the entry, so per (buffer, layout) a
+            // frame converts at most the largest range any draw asked for.
+            SizeT elementCount = 0;
+            // Pins the source buffer for the frame so its heap address cannot be reused by
+            // a new BufferObject while this pointer-keyed entry is alive.
+            SharedPtr<const MG_State::GLState::BufferObject> sourcePin;
+        };
+        UnorderedMap<ConvertedVertexStreamKey, ConvertedVertexStream, ConvertedVertexStreamKeyHash>
+            m_convertedVertexStreams;
 
         void CreateInstance();
         VkResult SetupDebugMessenger();
@@ -203,15 +529,30 @@ namespace MobileGL::MG_Backend::DirectVulkan {
         VkPipeline GetOrCreatePipeline(
             GLenum mode,
             const MG_State::GLState::ProgramObject& program,
+            const ProgramFactory::VkProgramObject& programObj,
+            ProgramFactory::CompileOptionFlags transformFlags,
             const MG_State::GLState::VertexArrayObject& vao,
             const RenderPassEntry& renderPassEntry);
+        VkPipeline GetOrCreateComputePipeline(const ProgramFactory::VkProgramObject& programObj);
+        void DestroyComputePipelines();
+        Bool PrepareStorageImageTextures(
+            VkCommandBuffer commandBuffer,
+            const MG_State::GLState::ProgramObject& program,
+            const ProgramFactory::VkProgramObject& programObj);
 
-        Bool UploadAndBindVertexBuffers(VkCommandBuffer commandBuffer, const MG_State::GLState::VertexArrayObject& vao);
+        Bool UploadAndBindVertexBuffers(VkCommandBuffer commandBuffer, const MG_State::GLState::VertexArrayObject& vao,
+                                        const ProgramFactory::VkProgramObject& programObj,
+                                        const DrawCmdParam& drawParams,
+                                        const IndexBufferView* pIndexBufferView);
         Bool UploadAndBindIndexBuffer(FrameContext::FrameData& frame,
                                      const MG_State::GLState::VertexArrayObject& vao,
                                       const IndexBufferView* pIndexBufferView = nullptr);
         Bool InitializeBlitResources();
+        Bool InitializeDepthMipmapResources();
         void ShutdownBlitResources();
+        void ShutdownDepthMipmapResources();
+        void CollectDeferredDepthMipmapCleanup(Uint32 frameIndex);
+        void DestroyDeferredDepthMipmapCleanup();
         Bool TryBlitToDefaultFramebufferWithShader(FrameContext::FrameData& frame,
                                                    MG_State::GLState::FramebufferObject& readFbo,
                                                    MG_State::GLState::FramebufferObject& drawFbo,
@@ -220,7 +561,19 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                                    GLenum filter);
         Bool MaterializePendingClearForTexture(VkCommandBuffer commandBuffer,
                                                MG_State::GLState::ITextureObject& texture);
+        Bool MaterializePendingClearForRenderbuffer(
+            VkCommandBuffer commandBuffer,
+            const SharedPtr<MG_State::GLState::RenderbufferObject>& renderbuffer);
         VkPipeline GetOrCreateBlitPipeline(const RenderPassEntry& renderPassEntry);
+        Bool GenerateDepthMipmapWithShader(FrameContext::FrameData& frame,
+                                           MG_State::GLState::ITextureObject& texture,
+                                           VkTextureManager::TextureResource& resource,
+                                           Uint32 baseMipLevel,
+                                           Uint32 generateMipLevelCount,
+                                           const IntVec3& storageBaseTexelSize,
+                                           VkImageLayout originalLayout,
+                                           VkImageLayout finalLayout);
+        Bool SubmitReadbackCommandsAndWait(FrameContext::FrameData& frame);
 
         void ShutdownSwapchain();
 

MobileGL/MG_Backend/DirectVulkan/VkIncludes.h

@@ -10,16 +10,61 @@
 
 #include "VulkanRendererConfig.h"
 
+#define ENUM_STR_CASE(c) case c: return #c;
+
+namespace MobileGL::MG_Backend::DirectVulkan {
+    inline const char* VkResultToString(VkResult result) {
+        switch (result) {
+        ENUM_STR_CASE(VK_SUCCESS)
+        ENUM_STR_CASE(VK_NOT_READY)
+        ENUM_STR_CASE(VK_TIMEOUT)
+        ENUM_STR_CASE(VK_EVENT_SET)
+        ENUM_STR_CASE(VK_EVENT_RESET)
+        ENUM_STR_CASE(VK_INCOMPLETE)
+        ENUM_STR_CASE(VK_ERROR_OUT_OF_HOST_MEMORY)
+        ENUM_STR_CASE(VK_ERROR_OUT_OF_DEVICE_MEMORY)
+        ENUM_STR_CASE(VK_ERROR_INITIALIZATION_FAILED)
+        ENUM_STR_CASE(VK_ERROR_DEVICE_LOST)
+        ENUM_STR_CASE(VK_ERROR_MEMORY_MAP_FAILED)
+        ENUM_STR_CASE(VK_ERROR_LAYER_NOT_PRESENT)
+        ENUM_STR_CASE(VK_ERROR_EXTENSION_NOT_PRESENT)
+        ENUM_STR_CASE(VK_ERROR_FEATURE_NOT_PRESENT)
+        ENUM_STR_CASE(VK_ERROR_INCOMPATIBLE_DRIVER)
+        ENUM_STR_CASE(VK_ERROR_TOO_MANY_OBJECTS)
+        ENUM_STR_CASE(VK_ERROR_FORMAT_NOT_SUPPORTED)
+        ENUM_STR_CASE(VK_ERROR_FRAGMENTED_POOL)
+        ENUM_STR_CASE(VK_ERROR_UNKNOWN)
+        ENUM_STR_CASE(VK_ERROR_OUT_OF_POOL_MEMORY)
+        ENUM_STR_CASE(VK_ERROR_INVALID_EXTERNAL_HANDLE)
+        ENUM_STR_CASE(VK_ERROR_FRAGMENTATION)
+        ENUM_STR_CASE(VK_ERROR_INVALID_OPAQUE_CAPTURE_ADDRESS)
+        ENUM_STR_CASE(VK_PIPELINE_COMPILE_REQUIRED)
+        ENUM_STR_CASE(VK_ERROR_SURFACE_LOST_KHR)
+        ENUM_STR_CASE(VK_ERROR_NATIVE_WINDOW_IN_USE_KHR)
+        ENUM_STR_CASE(VK_SUBOPTIMAL_KHR)
+        ENUM_STR_CASE(VK_ERROR_OUT_OF_DATE_KHR)
+        ENUM_STR_CASE(VK_ERROR_INCOMPATIBLE_DISPLAY_KHR)
+        ENUM_STR_CASE(VK_ERROR_VALIDATION_FAILED_EXT)
+        ENUM_STR_CASE(VK_ERROR_INVALID_SHADER_NV)
+        default:
+            return "VK_RESULT_UNKNOWN";
+        }
+    }
+} // namespace MobileGL::MG_Backend::DirectVulkan
+
 #define VK_VERIFY(expr, ...)                                                                                           \
     do {                                                                                                               \
         VkResult _vk_verify_result = (expr);                                                                           \
-        MOBILEGL_ASSERT(_vk_verify_result == VK_SUCCESS, "Vulkan error %d at %s:%d" __VA_OPT__(" - ") __VA_ARGS__, _vk_verify_result, __FILE__, __LINE__);  \
+        if (_vk_verify_result != VK_SUCCESS) {                                                                         \
+            MGLOG_F("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" __VA_OPT__(" - ") __VA_ARGS__, MobileGL::MG_Backend::DirectVulkan::VkResultToString(_vk_verify_result), _vk_verify_result, __FILE__, __LINE__);  \
     } while (0)
 
-#define ENUM_STR_CASE(c) case c: return #c;
-
 #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__);  \
-    } while (0)
+    } while (0)

MobileGL/MG_Backend/DirectVulkan/VulkanRendererConfig.h

@@ -11,10 +11,22 @@
 
 namespace MobileGL::MG_Backend::DirectVulkan {
     struct VulkanRendererConfig {
-        Uint32 MaxFramesInFlight = 2;
+        // Fallback CPU pipeline depth used when the MOBILEGL_MAGMA_FRAMESINFLIGHT env var is
+        // unset/invalid. A deeper pipeline lets the CPU run further ahead of the GPU, hiding
+        // per-frame GPU-completion latency. Whatever value is chosen (env or this fallback) is
+        // only a request: VulkanRenderer::Initialize clamps it down to the surface's maxImageCount
+        // (and never below 2), since not every driver allows that many swapchain images.
+        Uint32 MaxFramesInFlight = 3;
         String AppName = "MobileGL-VulkanRenderer";
         MobileGL::Version Version = MG_Config::CoreVersion;
         Uint64 CacheVersion = MG_Config::CacheVersion;
+        Uint32 SurfaceWidth = 1;
+        Uint32 SurfaceHeight = 1;
+        Bool DisablePipelineCache = false;
+#if MOBILEGL_LOG_ACTIVE_LEVEL <= MOBILEGL_LOG_LEVEL_DEBUG
         Bool EnableValidationLayers = true;
+#else
+        Bool EnableValidationLayers = false;
+#endif
     };
 } // namespace MobileGL::MG_Backend::DirectVulkan

MobileGL/MG_Benchmark/Buffer/CMakeLists.txt

@@ -16,4 +16,5 @@ target_link_libraries(
         ${LINK_LIBRARIES}
 )
 
-add_test(NAME BufferBench COMMAND BufferBench --benchmark_counters_tabular=true)
+add_test(NAME BufferBench COMMAND BufferBench --benchmark_counters_tabular=true)
+set_tests_properties(BufferBench PROPERTIES LABELS benchmark)

MobileGL/MG_Benchmark/CMakeLists.txt

@@ -38,6 +38,7 @@ target_link_libraries(
 )
 
 add_test(NAME SanityBench COMMAND SanityBench --benchmark_counters_tabular=true)
+set_tests_properties(SanityBench PROPERTIES LABELS benchmark)
 
 add_subdirectory(Program)
 add_subdirectory(Buffer)

MobileGL/MG_Benchmark/Program/CMakeLists.txt

@@ -16,4 +16,5 @@ target_link_libraries(
         ${LINK_LIBRARIES}
 )
 
-add_test(NAME ProgramBench COMMAND ProgramBench --benchmark_counters_tabular=true)
+add_test(NAME ProgramBench COMMAND ProgramBench --benchmark_counters_tabular=true)
+set_tests_properties(ProgramBench PROPERTIES LABELS benchmark)

MobileGL/MG_Impl/CGLImpl/CGLImpl.cpp

@@ -0,0 +1,700 @@
+// MobileGL - MobileGL/MG_Impl/CGLImpl/CGLImpl.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 "CGLImpl.h"
+
+#if defined(__APPLE__)
+#include "../EGLImpl/EGLImpl.h"
+
+namespace MobileGL::MG_Impl::CGLImpl {
+    namespace {
+        struct PixelFormatObject {
+            Uint32 RetainCount = 1;
+            Bool DoubleBuffer = true;
+            GLint ColorSize = 24;
+            GLint AlphaSize = 8;
+            GLint DepthSize = 24;
+            GLint StencilSize = 8;
+            GLint SampleBuffers = 0;
+            GLint Samples = 0;
+            GLint Profile = kCGLOGLPVersion_3_2_Core;
+            GLint RendererId = 0x4d474c;
+        };
+
+        struct ContextObject {
+            Uint32 RetainCount = 1;
+            CGLPixelFormatObj PixelFormat = nullptr;
+            CGLContextObj Share = nullptr;
+            EGLDisplay Display = EGL_NO_DISPLAY;
+            EGLConfig Config = nullptr;
+            EGLContext Context = EGL_NO_CONTEXT;
+            EGLSurface Surface = EGL_NO_SURFACE;
+            void* NSObject = nullptr;
+            void* View = nullptr;
+            void* MetalLayer = nullptr;
+            GLint SwapInterval = 1;
+            GLint VirtualScreen = 0;
+            GLint SurfaceBackingSize[2] = {0, 0};
+            Bool HasDrawable = false;
+            Bool Locked = false;
+        };
+
+        std::recursive_mutex& RegistryMutex() {
+            static auto* mutex = new std::recursive_mutex();
+            return *mutex;
+        }
+
+        Uint64& NextPixelFormatHandle() {
+            static auto* handle = new Uint64(1);
+            return *handle;
+        }
+
+        Uint64& NextContextHandle() {
+            static auto* handle = new Uint64(1);
+            return *handle;
+        }
+
+        UnorderedMap<CGLPixelFormatObj, PixelFormatObject>& PixelFormats() {
+            static auto* formats = new UnorderedMap<CGLPixelFormatObj, PixelFormatObject>();
+            return *formats;
+        }
+
+        UnorderedMap<CGLContextObj, ContextObject>& Contexts() {
+            static auto* contexts = new UnorderedMap<CGLContextObj, ContextObject>();
+            return *contexts;
+        }
+
+        UnorderedMap<std::thread::id, CGLContextObj>& CurrentContexts() {
+            static auto* contexts = new UnorderedMap<std::thread::id, CGLContextObj>();
+            return *contexts;
+        }
+
+        CGLPixelFormatObj EncodePixelFormat(Uint64 handle) {
+            return reinterpret_cast<CGLPixelFormatObj>(static_cast<SizeT>(handle));
+        }
+
+        CGLContextObj EncodeContext(Uint64 handle) {
+            return reinterpret_cast<CGLContextObj>(static_cast<SizeT>(handle));
+        }
+
+        std::thread::id CurrentThreadKey() {
+            return std::this_thread::get_id();
+        }
+
+        Bool AttributeHasValue(CGLPixelFormatAttribute attrib) {
+            switch (attrib) {
+            case kCGLPFAColorSize:
+            case kCGLPFAAlphaSize:
+            case kCGLPFADepthSize:
+            case kCGLPFAStencilSize:
+            case kCGLPFASampleBuffers:
+            case kCGLPFASamples:
+            case kCGLPFARendererID:
+            case kCGLPFADisplayMask:
+            case kCGLPFAOpenGLProfile:
+                return true;
+            default:
+                return false;
+            }
+        }
+
+        void ApplyPixelFormatAttribute(PixelFormatObject& pixelFormat,
+                                       CGLPixelFormatAttribute attrib,
+                                       GLint value) {
+            switch (attrib) {
+            case kCGLPFADoubleBuffer:
+                pixelFormat.DoubleBuffer = true;
+                break;
+            case kCGLPFAColorSize:
+                pixelFormat.ColorSize = value;
+                break;
+            case kCGLPFAAlphaSize:
+                pixelFormat.AlphaSize = value;
+                break;
+            case kCGLPFADepthSize:
+                pixelFormat.DepthSize = value;
+                break;
+            case kCGLPFAStencilSize:
+                pixelFormat.StencilSize = value;
+                break;
+            case kCGLPFASampleBuffers:
+                pixelFormat.SampleBuffers = value;
+                break;
+            case kCGLPFASamples:
+                pixelFormat.Samples = value;
+                break;
+            case kCGLPFAOpenGLProfile:
+                pixelFormat.Profile = value;
+                break;
+            case kCGLPFARendererID:
+                pixelFormat.RendererId = value;
+                break;
+            default:
+                break;
+            }
+        }
+
+        Bool InitEGLContext(ContextObject& object, CGLPixelFormatObj pix, CGLContextObj share) {
+            auto* pixelFormat = [&]() -> PixelFormatObject* {
+                auto& pixelFormats = PixelFormats();
+                auto it = pixelFormats.find(pix);
+                return it == pixelFormats.end() ? nullptr : &it->second;
+            }();
+            if (!pixelFormat) {
+                return false;
+            }
+
+            EGLDisplay display = EGLImpl::GetDisplay(EGL_DEFAULT_DISPLAY);
+            if (display == EGL_NO_DISPLAY) {
+                return false;
+            }
+            if (!EGLImpl::Initialize(display, nullptr, nullptr)) {
+                return false;
+            }
+            EGLImpl::BindAPI(EGL_OPENGL_API);
+
+            const EGLint attribs[] = {
+                EGL_RED_SIZE, 8,
+                EGL_GREEN_SIZE, 8,
+                EGL_BLUE_SIZE, 8,
+                EGL_ALPHA_SIZE, std::max(pixelFormat->AlphaSize, 0),
+                EGL_DEPTH_SIZE, std::max(pixelFormat->DepthSize, 0),
+                EGL_STENCIL_SIZE, std::max(pixelFormat->StencilSize, 0),
+                EGL_SURFACE_TYPE, EGL_WINDOW_BIT | EGL_PBUFFER_BIT,
+                EGL_RENDERABLE_TYPE, EGL_OPENGL_BIT,
+                EGL_NONE,
+            };
+
+            EGLConfig config = nullptr;
+            EGLint count = 0;
+            if (!EGLImpl::ChooseConfig(display, attribs, &config, 1, &count) || count <= 0) {
+                return false;
+            }
+
+            EGLContext shareContext = EGL_NO_CONTEXT;
+            if (share != nullptr) {
+                auto& contexts = Contexts();
+                auto shareIt = contexts.find(share);
+                if (shareIt == contexts.end()) {
+                    return false;
+                }
+                shareContext = shareIt->second.Context;
+            }
+
+            const EGLint contextAttribs[] = {
+                EGL_CONTEXT_MAJOR_VERSION, 3,
+                EGL_CONTEXT_MINOR_VERSION, 3,
+                EGL_NONE,
+            };
+            EGLContext eglContext = EGLImpl::CreateContext(display, config, shareContext, contextAttribs);
+            if (eglContext == EGL_NO_CONTEXT) {
+                return false;
+            }
+
+            object.Display = display;
+            object.Config = config;
+            object.Context = eglContext;
+            object.PixelFormat = pix;
+            object.Share = share;
+            return true;
+        }
+
+        ContextObject* TryGetContext(CGLContextObj ctx) {
+            auto& contexts = Contexts();
+            auto it = contexts.find(ctx);
+            return it == contexts.end() ? nullptr : &it->second;
+        }
+
+        const ContextObject* TryGetContext(CGLContextObj ctx, const std::lock_guard<std::recursive_mutex>&) {
+            auto& contexts = Contexts();
+            auto it = contexts.find(ctx);
+            return it == contexts.end() ? nullptr : &it->second;
+        }
+
+        PixelFormatObject* TryGetPixelFormat(CGLPixelFormatObj pix) {
+            auto& pixelFormats = PixelFormats();
+            auto it = pixelFormats.find(pix);
+            return it == pixelFormats.end() ? nullptr : &it->second;
+        }
+
+        CGLError MakeCurrentLocked(CGLContextObj ctx, ContextObject& object) {
+            CurrentContexts()[CurrentThreadKey()] = ctx;
+            if (!object.HasDrawable || object.Surface == EGL_NO_SURFACE) {
+                return kCGLNoError;
+            }
+            if (!EGLImpl::MakeCurrent(object.Display, object.Surface, object.Surface, object.Context)) {
+                return kCGLBadState;
+            }
+            return kCGLNoError;
+        }
+
+        CGLError RecreateSurfaceLocked(CGLContextObj ctx, ContextObject& object) {
+            if (!object.MetalLayer) {
+                return kCGLBadDrawable;
+            }
+            if (object.Surface != EGL_NO_SURFACE) {
+                EGLImpl::DestroySurface(object.Display, object.Surface);
+                object.Surface = EGL_NO_SURFACE;
+            }
+            const EGLAttrib attribs[] = {
+                EGL_WIDTH, std::max<GLint>(object.SurfaceBackingSize[0], 1),
+                EGL_HEIGHT, std::max<GLint>(object.SurfaceBackingSize[1], 1),
+                EGL_NONE,
+            };
+            EGLSurface surface = EGLImpl::CreatePlatformWindowSurface(object.Display, object.Config,
+                                                                      object.MetalLayer, attribs);
+            if (surface == EGL_NO_SURFACE) {
+                object.HasDrawable = false;
+                return kCGLBadDrawable;
+            }
+            object.Surface = surface;
+            object.HasDrawable = true;
+            return GetCurrentContext() == ctx ? MakeCurrentLocked(ctx, object) : kCGLNoError;
+        }
+
+        CGLError ResizeSurfaceLocked(ContextObject& object) {
+            if (object.Surface == EGL_NO_SURFACE) {
+                return kCGLBadDrawable;
+            }
+            return EGLImpl::ResizePlatformWindowSurface(
+                       object.Display, object.Surface,
+                       std::max<GLint>(object.SurfaceBackingSize[0], 1),
+                       std::max<GLint>(object.SurfaceBackingSize[1], 1))
+                       ? kCGLNoError
+                       : kCGLBadDrawable;
+        }
+    } // namespace
+
+    CGLError ChoosePixelFormat(const CGLPixelFormatAttribute* attribs, CGLPixelFormatObj* pix, GLint* npix) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        if (!pix || !npix) {
+            return kCGLBadAddress;
+        }
+
+        PixelFormatObject object;
+        if (attribs) {
+            for (SizeT i = 0; attribs[i] != static_cast<CGLPixelFormatAttribute>(0); ++i) {
+                const auto attrib = attribs[i];
+                GLint value = 1;
+                if (AttributeHasValue(attrib)) {
+                    value = static_cast<GLint>(attribs[++i]);
+                }
+                ApplyPixelFormatAttribute(object, attrib, value);
+            }
+        }
+
+        const auto handle = EncodePixelFormat(NextPixelFormatHandle()++);
+        PixelFormats()[handle] = object;
+        *pix = handle;
+        *npix = 1;
+        return kCGLNoError;
+    }
+
+    CGLError DestroyPixelFormat(CGLPixelFormatObj pix) {
+        ReleasePixelFormat(pix);
+        return kCGLNoError;
+    }
+
+    CGLError DescribePixelFormat(CGLPixelFormatObj pix, GLint pixNum, CGLPixelFormatAttribute attrib, GLint* value) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        if (!value) {
+            return kCGLBadAddress;
+        }
+        if (pixNum != 0 && pixNum != 1) {
+            return kCGLBadValue;
+        }
+        auto* pixelFormat = TryGetPixelFormat(pix);
+        if (!pixelFormat) {
+            return kCGLBadPixelFormat;
+        }
+
+        switch (attrib) {
+        case kCGLPFADoubleBuffer:
+            *value = pixelFormat->DoubleBuffer ? 1 : 0;
+            return kCGLNoError;
+        case kCGLPFAAccelerated:
+        case kCGLPFAAcceleratedCompute:
+        case kCGLPFASupportsAutomaticGraphicsSwitching:
+            *value = 1;
+            return kCGLNoError;
+        case kCGLPFAColorSize:
+            *value = pixelFormat->ColorSize;
+            return kCGLNoError;
+        case kCGLPFAAlphaSize:
+            *value = pixelFormat->AlphaSize;
+            return kCGLNoError;
+        case kCGLPFADepthSize:
+            *value = pixelFormat->DepthSize;
+            return kCGLNoError;
+        case kCGLPFAStencilSize:
+            *value = pixelFormat->StencilSize;
+            return kCGLNoError;
+        case kCGLPFASampleBuffers:
+            *value = pixelFormat->SampleBuffers;
+            return kCGLNoError;
+        case kCGLPFASamples:
+            *value = pixelFormat->Samples;
+            return kCGLNoError;
+        case kCGLPFARendererID:
+            *value = pixelFormat->RendererId;
+            return kCGLNoError;
+        case kCGLPFAOpenGLProfile:
+            *value = pixelFormat->Profile;
+            return kCGLNoError;
+        case kCGLPFAVirtualScreenCount:
+            *value = 1;
+            return kCGLNoError;
+        default:
+            *value = 0;
+            return kCGLNoError;
+        }
+    }
+
+    void ReleasePixelFormat(CGLPixelFormatObj pix) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* pixelFormat = TryGetPixelFormat(pix);
+        if (!pixelFormat) {
+            return;
+        }
+        if (pixelFormat->RetainCount > 1) {
+            --pixelFormat->RetainCount;
+            return;
+        }
+        PixelFormats().erase(pix);
+    }
+
+    CGLPixelFormatObj RetainPixelFormat(CGLPixelFormatObj pix) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* pixelFormat = TryGetPixelFormat(pix);
+        if (pixelFormat) {
+            ++pixelFormat->RetainCount;
+        }
+        return pix;
+    }
+
+    GLuint GetPixelFormatRetainCount(CGLPixelFormatObj pix) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* pixelFormat = TryGetPixelFormat(pix);
+        return pixelFormat ? pixelFormat->RetainCount : 0;
+    }
+
+    CGLError CreateContext(CGLPixelFormatObj pix, CGLContextObj share, CGLContextObj* ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        if (!ctx) {
+            return kCGLBadAddress;
+        }
+        if (!TryGetPixelFormat(pix)) {
+            return kCGLBadPixelFormat;
+        }
+        if (share && !TryGetContext(share)) {
+            return kCGLBadMatch;
+        }
+
+        ContextObject object;
+        if (!InitEGLContext(object, pix, share)) {
+            return kCGLBadAlloc;
+        }
+        RetainPixelFormat(pix);
+        const auto handle = EncodeContext(NextContextHandle()++);
+        Contexts()[handle] = object;
+        *ctx = handle;
+        return kCGLNoError;
+    }
+
+    CGLError DestroyContext(CGLContextObj ctx) {
+        ReleaseContext(ctx);
+        return kCGLNoError;
+    }
+
+    CGLContextObj RetainContext(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (object) {
+            ++object->RetainCount;
+        }
+        return ctx;
+    }
+
+    void ReleaseContext(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return;
+        }
+        if (object->RetainCount > 1) {
+            --object->RetainCount;
+            return;
+        }
+        if (object->Surface != EGL_NO_SURFACE) {
+            EGLImpl::DestroySurface(object->Display, object->Surface);
+        }
+        if (object->Context != EGL_NO_CONTEXT) {
+            EGLImpl::DestroyContext(object->Display, object->Context);
+        }
+        ReleasePixelFormat(object->PixelFormat);
+        auto& currentContexts = CurrentContexts();
+        for (auto it = currentContexts.begin(); it != currentContexts.end();) {
+            if (it->second == ctx) {
+                it = currentContexts.erase(it);
+            } else {
+                ++it;
+            }
+        }
+        Contexts().erase(ctx);
+    }
+
+    GLuint GetContextRetainCount(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        return object ? object->RetainCount : 0;
+    }
+
+    CGLPixelFormatObj GetPixelFormat(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        return object ? object->PixelFormat : nullptr;
+    }
+
+    CGLError SetCurrentContext(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        if (!ctx) {
+            CurrentContexts().erase(CurrentThreadKey());
+            EGLImpl::MakeCurrent(EGL_NO_DISPLAY, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
+            return kCGLNoError;
+        }
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        return MakeCurrentLocked(ctx, *object);
+    }
+
+    CGLContextObj GetCurrentContext() {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto& currentContexts = CurrentContexts();
+        auto it = currentContexts.find(CurrentThreadKey());
+        return it == currentContexts.end() ? nullptr : it->second;
+    }
+
+    CGLError SetParameter(CGLContextObj ctx, CGLContextParameter pname, const GLint* params) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        if (!params && pname != kCGLCPReclaimResources) {
+            return kCGLBadAddress;
+        }
+        switch (pname) {
+        case kCGLCPSwapInterval:
+            object->SwapInterval = params[0];
+            EGLImpl::SwapInterval(object->Display, object->SwapInterval);
+            return kCGLNoError;
+        case kCGLCPSurfaceBackingSize:
+        {
+            const GLint width = std::max<GLint>(params[0], 1);
+            const GLint height = std::max<GLint>(params[1], 1);
+            if (object->SurfaceBackingSize[0] == width && object->SurfaceBackingSize[1] == height) {
+                return kCGLNoError;
+            }
+            object->SurfaceBackingSize[0] = width;
+            object->SurfaceBackingSize[1] = height;
+            if (object->MetalLayer && object->Surface != EGL_NO_SURFACE) {
+                return ResizeSurfaceLocked(*object);
+            }
+            return kCGLNoError;
+        }
+        case kCGLCPSurfaceOpacity:
+        case kCGLCPSurfaceOrder:
+        case kCGLCPMPSwapsInFlight:
+        case kCGLCPReclaimResources:
+            return kCGLNoError;
+        default:
+            return kCGLNoError;
+        }
+    }
+
+    CGLError GetParameter(CGLContextObj ctx, CGLContextParameter pname, GLint* params) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        if (!params) {
+            return kCGLBadAddress;
+        }
+        switch (pname) {
+        case kCGLCPSwapInterval:
+            params[0] = object->SwapInterval;
+            return kCGLNoError;
+        case kCGLCPSurfaceBackingSize:
+            params[0] = object->SurfaceBackingSize[0];
+            params[1] = object->SurfaceBackingSize[1];
+            return kCGLNoError;
+        case kCGLCPCurrentRendererID:
+            params[0] = 0x4d474c;
+            return kCGLNoError;
+        case kCGLCPGPUVertexProcessing:
+        case kCGLCPGPUFragmentProcessing:
+        case kCGLCPHasDrawable:
+            params[0] = object->HasDrawable ? 1 : 0;
+            return kCGLNoError;
+        case kCGLCPMPSwapsInFlight:
+            params[0] = 1;
+            return kCGLNoError;
+        default:
+            params[0] = 0;
+            return kCGLNoError;
+        }
+    }
+
+    CGLError UpdateContext(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        return TryGetContext(ctx) ? kCGLNoError : kCGLBadContext;
+    }
+
+    CGLError ClearDrawable(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        if (object->Surface != EGL_NO_SURFACE) {
+            EGLImpl::DestroySurface(object->Display, object->Surface);
+        }
+        object->Surface = EGL_NO_SURFACE;
+        object->View = nullptr;
+        object->MetalLayer = nullptr;
+        object->HasDrawable = false;
+        return kCGLNoError;
+    }
+
+    CGLError FlushDrawable(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        if (!object->HasDrawable || object->Surface == EGL_NO_SURFACE) {
+            return kCGLBadDrawable;
+        }
+        const auto currentError = MakeCurrentLocked(ctx, *object);
+        if (currentError != kCGLNoError) {
+            return currentError;
+        }
+        return EGLImpl::SwapBuffers(object->Display, object->Surface) ? kCGLNoError : kCGLBadDrawable;
+    }
+
+    CGLError LockContext(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        object->Locked = true;
+        return kCGLNoError;
+    }
+
+    CGLError UnlockContext(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        object->Locked = false;
+        return kCGLNoError;
+    }
+
+    void GetVersion(GLint* majorvers, GLint* minorvers) {
+        if (majorvers) {
+            *majorvers = 1;
+        }
+        if (minorvers) {
+            *minorvers = 0;
+        }
+    }
+
+    const char* ErrorString(CGLError error) {
+        switch (error) {
+        case kCGLNoError:
+            return "no error";
+        case kCGLBadAttribute:
+            return "invalid pixel format attribute";
+        case kCGLBadPixelFormat:
+            return "invalid pixel format";
+        case kCGLBadContext:
+            return "invalid context";
+        case kCGLBadDrawable:
+            return "invalid drawable";
+        case kCGLBadState:
+            return "invalid context state";
+        case kCGLBadValue:
+            return "invalid numerical value";
+        case kCGLBadMatch:
+            return "invalid share context";
+        case kCGLBadAddress:
+            return "invalid pointer";
+        case kCGLBadAlloc:
+            return "invalid memory allocation";
+        default:
+            return "unknown CGL error";
+        }
+    }
+
+    CGLError AttachDrawable(CGLContextObj ctx, void* nsView, void* metalLayer, GLint width, GLint height) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (!object) {
+            return kCGLBadContext;
+        }
+        if (!metalLayer) {
+            return kCGLBadDrawable;
+        }
+        width = std::max<GLint>(width, 1);
+        height = std::max<GLint>(height, 1);
+        const Bool sameSize = object->SurfaceBackingSize[0] == width && object->SurfaceBackingSize[1] == height;
+        object->SurfaceBackingSize[0] = width;
+        object->SurfaceBackingSize[1] = height;
+        if (object->Surface != EGL_NO_SURFACE && object->MetalLayer == metalLayer && sameSize) {
+            object->View = nsView;
+            object->HasDrawable = true;
+            return kCGLNoError;
+        }
+        if (object->Surface != EGL_NO_SURFACE && object->MetalLayer == metalLayer) {
+            object->View = nsView;
+            object->HasDrawable = true;
+            return ResizeSurfaceLocked(*object);
+        }
+        if (object->Surface != EGL_NO_SURFACE) {
+            EGLImpl::DestroySurface(object->Display, object->Surface);
+            object->Surface = EGL_NO_SURFACE;
+        }
+        object->View = nsView;
+        object->MetalLayer = metalLayer;
+        const auto recreateError = RecreateSurfaceLocked(ctx, *object);
+        if (recreateError != kCGLNoError) {
+            return recreateError;
+        }
+        return kCGLNoError;
+    }
+
+    void* GetContextNSObject(CGLContextObj ctx) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        return object ? object->NSObject : nullptr;
+    }
+
+    void SetContextNSObject(CGLContextObj ctx, void* nsObject) {
+        const std::lock_guard<std::recursive_mutex> lock(RegistryMutex());
+        auto* object = TryGetContext(ctx);
+        if (object) {
+            object->NSObject = nsObject;
+        }
+    }
+} // namespace MobileGL::MG_Impl::CGLImpl
+#endif

MobileGL/MG_Impl/CGLImpl/CGLImpl.h

@@ -0,0 +1,49 @@
+// MobileGL - MobileGL/MG_Impl/CGLImpl/CGLImpl.h
+// Copyright (c) 2025-2026 MobileGL-Dev
+// Licensed under the GNU Lesser General Public License v3.0:
+//   https://www.gnu.org/licenses/gpl-3.0.txt
+//   https://www.gnu.org/licenses/lgpl-3.0.txt
+// SPDX-License-Identifier: LGPL-3.0-only
+// End of Source File Header
+
+#pragma once
+#include <Includes.h>
+
+#if defined(__APPLE__)
+#ifndef GL_SILENCE_DEPRECATION
+#define GL_SILENCE_DEPRECATION
+#endif
+#include <OpenGL/OpenGL.h>
+
+namespace MobileGL::MG_Impl::CGLImpl {
+    CGLError ChoosePixelFormat(const CGLPixelFormatAttribute* attribs, CGLPixelFormatObj* pix, GLint* npix);
+    CGLError DestroyPixelFormat(CGLPixelFormatObj pix);
+    CGLError DescribePixelFormat(CGLPixelFormatObj pix, GLint pixNum, CGLPixelFormatAttribute attrib, GLint* value);
+    void ReleasePixelFormat(CGLPixelFormatObj pix);
+    CGLPixelFormatObj RetainPixelFormat(CGLPixelFormatObj pix);
+    GLuint GetPixelFormatRetainCount(CGLPixelFormatObj pix);
+
+    CGLError CreateContext(CGLPixelFormatObj pix, CGLContextObj share, CGLContextObj* ctx);
+    CGLError DestroyContext(CGLContextObj ctx);
+    CGLContextObj RetainContext(CGLContextObj ctx);
+    void ReleaseContext(CGLContextObj ctx);
+    GLuint GetContextRetainCount(CGLContextObj ctx);
+    CGLPixelFormatObj GetPixelFormat(CGLContextObj ctx);
+
+    CGLError SetCurrentContext(CGLContextObj ctx);
+    CGLContextObj GetCurrentContext();
+    CGLError SetParameter(CGLContextObj ctx, CGLContextParameter pname, const GLint* params);
+    CGLError GetParameter(CGLContextObj ctx, CGLContextParameter pname, GLint* params);
+    CGLError UpdateContext(CGLContextObj ctx);
+    CGLError ClearDrawable(CGLContextObj ctx);
+    CGLError FlushDrawable(CGLContextObj ctx);
+    CGLError LockContext(CGLContextObj ctx);
+    CGLError UnlockContext(CGLContextObj ctx);
+    void GetVersion(GLint* majorvers, GLint* minorvers);
+    const char* ErrorString(CGLError error);
+
+    CGLError AttachDrawable(CGLContextObj ctx, void* nsView, void* metalLayer, GLint width, GLint height);
+    void* GetContextNSObject(CGLContextObj ctx);
+    void SetContextNSObject(CGLContextObj ctx, void* nsObject);
+}
+#endif

MobileGL/MG_Impl/CGLImpl/Exporting/Definitions.cpp

@@ -0,0 +1,110 @@
+// MobileGL - MobileGL/MG_Impl/CGLImpl/Exporting/Definitions.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 "../CGLImpl.h"
+
+#if defined(__APPLE__)
+
+MOBILEGL_CGL_API CGLError CGLChoosePixelFormat(const CGLPixelFormatAttribute* attribs,
+                                               CGLPixelFormatObj* pix,
+                                               GLint* npix) {
+    return MobileGL::MG_Impl::CGLImpl::ChoosePixelFormat(attribs, pix, npix);
+}
+
+MOBILEGL_CGL_API CGLError CGLDestroyPixelFormat(CGLPixelFormatObj pix) {
+    return MobileGL::MG_Impl::CGLImpl::DestroyPixelFormat(pix);
+}
+
+MOBILEGL_CGL_API CGLError CGLDescribePixelFormat(CGLPixelFormatObj pix,
+                                                 GLint pix_num,
+                                                 CGLPixelFormatAttribute attrib,
+                                                 GLint* value) {
+    return MobileGL::MG_Impl::CGLImpl::DescribePixelFormat(pix, pix_num, attrib, value);
+}
+
+MOBILEGL_CGL_API void CGLReleasePixelFormat(CGLPixelFormatObj pix) {
+    MobileGL::MG_Impl::CGLImpl::ReleasePixelFormat(pix);
+}
+
+MOBILEGL_CGL_API CGLPixelFormatObj CGLRetainPixelFormat(CGLPixelFormatObj pix) {
+    return MobileGL::MG_Impl::CGLImpl::RetainPixelFormat(pix);
+}
+
+MOBILEGL_CGL_API GLuint CGLGetPixelFormatRetainCount(CGLPixelFormatObj pix) {
+    return MobileGL::MG_Impl::CGLImpl::GetPixelFormatRetainCount(pix);
+}
+
+MOBILEGL_CGL_API CGLError CGLCreateContext(CGLPixelFormatObj pix, CGLContextObj share, CGLContextObj* ctx) {
+    return MobileGL::MG_Impl::CGLImpl::CreateContext(pix, share, ctx);
+}
+
+MOBILEGL_CGL_API CGLError CGLDestroyContext(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::DestroyContext(ctx);
+}
+
+MOBILEGL_CGL_API CGLContextObj CGLRetainContext(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::RetainContext(ctx);
+}
+
+MOBILEGL_CGL_API void CGLReleaseContext(CGLContextObj ctx) {
+    MobileGL::MG_Impl::CGLImpl::ReleaseContext(ctx);
+}
+
+MOBILEGL_CGL_API GLuint CGLGetContextRetainCount(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::GetContextRetainCount(ctx);
+}
+
+MOBILEGL_CGL_API CGLPixelFormatObj CGLGetPixelFormat(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::GetPixelFormat(ctx);
+}
+
+MOBILEGL_CGL_API CGLError CGLSetCurrentContext(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::SetCurrentContext(ctx);
+}
+
+MOBILEGL_CGL_API CGLContextObj CGLGetCurrentContext(void) {
+    return MobileGL::MG_Impl::CGLImpl::GetCurrentContext();
+}
+
+MOBILEGL_CGL_API CGLError CGLSetParameter(CGLContextObj ctx, CGLContextParameter pname, const GLint* params) {
+    return MobileGL::MG_Impl::CGLImpl::SetParameter(ctx, pname, params);
+}
+
+MOBILEGL_CGL_API CGLError CGLGetParameter(CGLContextObj ctx, CGLContextParameter pname, GLint* params) {
+    return MobileGL::MG_Impl::CGLImpl::GetParameter(ctx, pname, params);
+}
+
+MOBILEGL_CGL_API CGLError CGLUpdateContext(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::UpdateContext(ctx);
+}
+
+MOBILEGL_CGL_API CGLError CGLClearDrawable(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::ClearDrawable(ctx);
+}
+
+MOBILEGL_CGL_API CGLError CGLFlushDrawable(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::FlushDrawable(ctx);
+}
+
+MOBILEGL_CGL_API CGLError CGLLockContext(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::LockContext(ctx);
+}
+
+MOBILEGL_CGL_API CGLError CGLUnlockContext(CGLContextObj ctx) {
+    return MobileGL::MG_Impl::CGLImpl::UnlockContext(ctx);
+}
+
+MOBILEGL_CGL_API void CGLGetVersion(GLint* majorvers, GLint* minorvers) {
+    MobileGL::MG_Impl::CGLImpl::GetVersion(majorvers, minorvers);
+}
+
+MOBILEGL_CGL_API const char* CGLErrorString(CGLError error) {
+    return MobileGL::MG_Impl::CGLImpl::ErrorString(error);
+}
+
+#endif

MobileGL/MG_Impl/DyldInterpose/DyldInterpose.cpp

@@ -0,0 +1,56 @@
+// MobileGL - MobileGL/MG_Impl/DyldInterpose/DyldInterpose.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 <Includes.h>
+
+#if defined(__APPLE__)
+
+#include "MG_Impl/GetProcAddress.h"
+
+#include <dlfcn.h>
+
+namespace {
+    struct DyldInterposeEntry {
+        const void* Replacement;
+        const void* Replacee;
+    };
+
+    bool IsGLProcName(const char* name) {
+        if (name == nullptr) {
+            return false;
+        }
+
+        if (strncmp(name, "CGL", 3) == 0) {
+            return true;
+        }
+
+        if (strncmp(name, "gl", 2) != 0) {
+            return false;
+        }
+
+        // Avoid stealing glfw*/glib*/glX*/global application symbols.
+        return name[2] >= 'A' && name[2] <= 'Z' && name[2] != 'X';
+    }
+
+    void* MobileGLDlsym(void* handle, const char* symbol) {
+        if (IsGLProcName(symbol)) {
+            if (void* proc = MobileGL::MG_Impl::GetProcAddress(symbol)) {
+                return proc;
+            }
+        }
+
+        return dlsym(handle, symbol);
+    }
+
+    __attribute__((used)) static const DyldInterposeEntry kMobileGLDyldInterpose[]
+        __attribute__((section("__DATA,__interpose"))) = {
+            {reinterpret_cast<const void*>(MobileGLDlsym), reinterpret_cast<const void*>(dlsym)},
+        };
+} // namespace
+
+#endif

MobileGL/MG_Impl/EGLImpl/EGLImpl.cpp

@@ -8,8 +8,11 @@
 
 #include "EGLImpl.h"
 #include "../GetProcAddress.h"
+#include <Init.h>
 #include <MG_Backend/BackendObjects.h>
 #include <MG_State/EGLState/Core.h>
+#include <mutex>
+#include <sstream>
 #include <type_traits>
 
 namespace MobileGL::MG_Impl::EGLImpl {
@@ -23,6 +26,17 @@ namespace MobileGL::MG_Impl::EGLImpl {
             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) {
@@ -31,14 +45,55 @@ namespace MobileGL::MG_Impl::EGLImpl {
             return backendObject;
         }
 
+        std::recursive_mutex& EGLOperationMutex() {
+            static std::recursive_mutex mutex;
+            return mutex;
+        }
+
+        String CurrentThreadIdString() {
+            std::ostringstream stream;
+            stream << std::this_thread::get_id();
+            return stream.str();
+        }
+
         MG_Backend::WindowBackend DetectWindowBackend() {
 #if defined(ANDROID) || defined(__ANDROID__)
             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
             return MG_Backend::WindowBackend::Unknown;
 #endif
         }
 
+        EGLint GetAttribValue(const EGLint* attribList, EGLint attrib, EGLint defaultValue) {
+            if (!attribList) {
+                return defaultValue;
+            }
+            for (SizeT i = 0; attribList[i] != EGL_NONE; i += 2) {
+                if (attribList[i] == attrib) {
+                    return attribList[i + 1];
+                }
+            }
+            return defaultValue;
+        }
+
+        EGLint GetAttribValueAttrib(const EGLAttrib* attribList, EGLint attrib, EGLint defaultValue) {
+            if (!attribList) {
+                return defaultValue;
+            }
+            for (SizeT i = 0; attribList[i] != EGL_NONE; i += 2) {
+                if (attribList[i] == attrib) {
+                    return static_cast<EGLint>(attribList[i + 1]);
+                }
+            }
+            return defaultValue;
+        }
+
         template <typename NativeType>
         Bool IsNullNativeHandle(NativeType nativeHandle) {
             if constexpr (std::is_pointer_v<NativeType>) {
@@ -77,25 +132,35 @@ namespace MobileGL::MG_Impl::EGLImpl {
             return EGL_NO_SURFACE;
         }
 
-        auto* backendObject = GetBackendObject(state);
-        if (!backendObject) {
-            MGLOG_E("activeBackendObject not initialized!");
-            return EGL_NO_SURFACE;
-        }
-
         const MG_Backend::WindowHandle windowHandle = {
             .Backend = DetectWindowBackend(),
             .Handle = ToVoidHandle(window),
+            .Width = static_cast<Uint32>(std::max<EGLint>(GetAttribValue(attrib_list, EGL_WIDTH, 0), 0)),
+            .Height = static_cast<Uint32>(std::max<EGLint>(GetAttribValue(attrib_list, EGL_HEIGHT, 0), 0)),
         };
-        if (!backendObject->CreateEGLWindowSurface(windowHandle)) {
+
+        EGLSurface surface = state->CreateWindowSurface(dpy, config, window, attrib_list);
+        if (surface == EGL_NO_SURFACE) {
+            return EGL_NO_SURFACE;
+        }
+
+        auto* backendObject = GetBackendObject(state);
+        if (!backendObject) {
+            MGLOG_E("activeBackendObject not initialized!");
+            state->DestroySurface(dpy, surface);
+            return EGL_NO_SURFACE;
+        }
+        if (!backendObject->CreateEGLWindowSurface(surface, windowHandle)) {
+            state->DestroySurface(dpy, surface);
             state->SetError(EGL_BAD_NATIVE_WINDOW);
             return EGL_NO_SURFACE;
         }
 
-        return state->CreateWindowSurface(dpy, config, window, attrib_list);
+        return surface;
     }
 
     EGLBoolean SwapBuffers(EGLDisplay dpy, EGLSurface draw) {
+        const std::lock_guard<std::recursive_mutex> operationLock(EGLOperationMutex());
         auto* state = GetState();
         if (!state) {
             return EGL_FALSE;
@@ -111,6 +176,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
             return EGL_FALSE;
         }
         if (!backendObject->SwapEGLBuffers(dpy, draw)) {
+            MGLOG_E("eglSwapBuffers failed on thread=%s dpy=%p draw=%p", CurrentThreadIdString().c_str(), dpy, draw);
             state->SetError(EGL_BAD_SURFACE);
             return EGL_FALSE;
         }
@@ -135,7 +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;
         }
@@ -156,7 +222,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     EGLDisplay GetDisplay(NativeDisplayType display) {
-        auto* state = GetState();
+        auto* state = GetStateEnsureInitialized();
         if (!state) {
             return EGL_NO_DISPLAY;
         }
@@ -172,6 +238,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     EGLBoolean MakeCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx) {
+        const std::lock_guard<std::recursive_mutex> operationLock(EGLOperationMutex());
         auto* state = GetState();
         if (!state) {
             return EGL_FALSE;
@@ -181,17 +248,30 @@ namespace MobileGL::MG_Impl::EGLImpl {
         const auto oldDraw = state->GetCurrentSurface(EGL_DRAW);
         const auto oldRead = state->GetCurrentSurface(EGL_READ);
         const auto oldContext = state->GetCurrentContext();
+        const String threadId = CurrentThreadIdString();
+
+        MGLOG_D("eglMakeCurrent begin thread=%s dpy=%p draw=%p read=%p ctx=%p oldDpy=%p oldDraw=%p oldRead=%p oldCtx=%p",
+                threadId.c_str(), dpy, draw, read, ctx, oldDisplay, oldDraw, oldRead, oldContext);
 
         if (!state->MakeCurrent(dpy, draw, read, ctx)) {
+            const EGLint error = state->ConsumeError();
+            MGLOG_D("eglMakeCurrent rejected by EGLState thread=%s error=0x%04x", threadId.c_str(), error);
+            state->SetError(error);
             return EGL_FALSE;
         }
 
         const Bool releaseCurrentRequest =
-            dpy == EGL_NO_DISPLAY && draw == EGL_NO_SURFACE && read == EGL_NO_SURFACE && ctx == EGL_NO_CONTEXT;
+            draw == EGL_NO_SURFACE && read == EGL_NO_SURFACE && ctx == EGL_NO_CONTEXT;
         if (releaseCurrentRequest) {
             if (auto* backendObject = MG_Backend::pActiveBackendObject.get()) {
-                (void)backendObject->MakeEGLCurrent(dpy, draw, read, ctx);
+                if (!backendObject->MakeEGLCurrent(dpy, draw, read, ctx)) {
+                    MGLOG_E("eglMakeCurrent release failed in backend thread=%s", threadId.c_str());
+                    state->MakeCurrent(oldDisplay, oldDraw, oldRead, oldContext);
+                    state->SetError(EGL_BAD_ACCESS);
+                    return EGL_FALSE;
+                }
             }
+            MGLOG_D("eglMakeCurrent release succeeded thread=%s", threadId.c_str());
             return EGL_TRUE;
         }
 
@@ -202,10 +282,14 @@ namespace MobileGL::MG_Impl::EGLImpl {
             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(),
+                    dpy, draw, read, ctx);
             state->SetError(EGL_BAD_ACCESS);
             state->MakeCurrent(oldDisplay, oldDraw, oldRead, oldContext);
             return EGL_FALSE;
         }
+        MGLOG_D("eglMakeCurrent attach succeeded thread=%s dpy=%p draw=%p read=%p ctx=%p", threadId.c_str(), dpy, draw,
+                read, ctx);
         return EGL_TRUE;
     }
 
@@ -218,11 +302,18 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     EGLBoolean DestroySurface(EGLDisplay dpy, EGLSurface surface) {
+        const std::lock_guard<std::recursive_mutex> operationLock(EGLOperationMutex());
         auto* state = GetState();
         if (!state) {
             return EGL_FALSE;
         }
-        return state->DestroySurface(dpy, surface) ? EGL_TRUE : EGL_FALSE;
+        if (!state->DestroySurface(dpy, surface)) {
+            return EGL_FALSE;
+        }
+        if (auto* backendObject = MG_Backend::pActiveBackendObject.get()) {
+            backendObject->ReleaseEGLSurface(surface);
+        }
+        return EGL_TRUE;
     }
 
     EGLBoolean Terminate(EGLDisplay dpy) {
@@ -230,7 +321,21 @@ namespace MobileGL::MG_Impl::EGLImpl {
         if (!state) {
             return EGL_FALSE;
         }
-        return state->TerminateDisplay(dpy) ? EGL_TRUE : EGL_FALSE;
+        if (!state->TerminateDisplay(dpy)) {
+            return EGL_FALSE;
+        }
+        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;
     }
 
     EGLBoolean ReleaseThread() {
@@ -238,6 +343,9 @@ namespace MobileGL::MG_Impl::EGLImpl {
         if (!state) {
             return EGL_FALSE;
         }
+        if (auto* backendObject = MG_Backend::pActiveBackendObject.get()) {
+            (void)backendObject->MakeEGLCurrent(EGL_NO_DISPLAY, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
+        }
         state->ReleaseThread();
         return EGL_TRUE;
     }
@@ -259,7 +367,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     EGLBoolean BindAPI(EGLenum api) {
-        auto* state = GetState();
+        auto* state = GetStateEnsureInitialized();
         if (!state) {
             return EGL_FALSE;
         }
@@ -292,7 +400,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     }
 
     char const* QueryString(EGLDisplay display, EGLint name) {
-        auto* state = GetState();
+        auto* state = GetStateEnsureInitialized();
         if (!state) {
             return nullptr;
         }
@@ -310,7 +418,14 @@ namespace MobileGL::MG_Impl::EGLImpl {
         case EGL_CLIENT_APIS:
             return "OpenGL OpenGL_ES";
         case EGL_EXTENSIONS:
-            return "";
+            if (display == EGL_NO_DISPLAY) {
+                return "EGL_EXT_client_extensions "
+                       "EGL_EXT_platform_base "
+                       "EGL_KHR_platform_base "
+                       "EGL_MESA_platform_surfaceless";
+            }
+            return "EGL_KHR_create_context "
+                   "EGL_MESA_platform_surfaceless";
         default:
             state->SetError(EGL_BAD_PARAMETER);
             return nullptr;
@@ -322,7 +437,17 @@ namespace MobileGL::MG_Impl::EGLImpl {
         if (!state) {
             return EGL_FALSE;
         }
-        return state->SwapInterval(dpy, interval) ? EGL_TRUE : EGL_FALSE;
+        if (!state->SwapInterval(dpy, interval)) {
+            return EGL_FALSE;
+        }
+        // Forward the request to the backend's native presentation path; without this
+        // the app's vsync setting only ever reaches MobileGL's shadow state and the
+        // native surface stays at the driver default (interval 1 = always vsynced).
+        auto* backendObject = GetBackendObject(state);
+        if (backendObject) {
+            backendObject->SetEGLSwapInterval(static_cast<Int>(interval));
+        }
+        return EGL_TRUE;
     }
 
     EGLSurface CreatePbufferSurface(EGLDisplay dpy, EGLConfig config, const EGLint* attrib_list) {
@@ -330,7 +455,24 @@ namespace MobileGL::MG_Impl::EGLImpl {
         if (!state) {
             return EGL_NO_SURFACE;
         }
-        return state->CreatePbufferSurface(dpy, config, attrib_list);
+        const EGLint width = GetAttribValue(attrib_list, EGL_WIDTH, 1);
+        const EGLint height = GetAttribValue(attrib_list, EGL_HEIGHT, 1);
+        EGLSurface surface = state->CreatePbufferSurface(dpy, config, attrib_list);
+        if (surface == EGL_NO_SURFACE) {
+            return EGL_NO_SURFACE;
+        }
+
+        auto* backendObject = GetBackendObject(state);
+        if (!backendObject) {
+            return EGL_NO_SURFACE;
+        }
+        if (!backendObject->CreateEGLPbufferSurface(surface, width, height)) {
+            state->DestroySurface(dpy, surface);
+            state->SetError(EGL_BAD_ALLOC);
+            return EGL_NO_SURFACE;
+        }
+
+        return surface;
     }
 
     EGLBoolean BindTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer) {
@@ -521,7 +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;
         }
@@ -547,22 +689,53 @@ namespace MobileGL::MG_Impl::EGLImpl {
             return EGL_NO_SURFACE;
         }
 
-        auto* backendObject = GetBackendObject(state);
-        if (!backendObject) {
-            MGLOG_E("activeBackendObject not initialized!");
-            return EGL_NO_SURFACE;
-        }
-
         const MG_Backend::WindowHandle windowHandle = {
             .Backend = DetectWindowBackend(),
             .Handle = native_window,
+            .Width = static_cast<Uint32>(std::max<EGLint>(GetAttribValueAttrib(attrib_list, EGL_WIDTH, 0), 0)),
+            .Height = static_cast<Uint32>(std::max<EGLint>(GetAttribValueAttrib(attrib_list, EGL_HEIGHT, 0), 0)),
         };
-        if (!backendObject->CreateEGLWindowSurface(windowHandle)) {
+
+        EGLSurface surface = state->CreatePlatformWindowSurface(dpy, config, native_window, attrib_list);
+        if (surface == EGL_NO_SURFACE) {
+            return EGL_NO_SURFACE;
+        }
+
+        auto* backendObject = GetBackendObject(state);
+        if (!backendObject) {
+            MGLOG_E("activeBackendObject not initialized!");
+            state->DestroySurface(dpy, surface);
+            return EGL_NO_SURFACE;
+        }
+        if (!backendObject->CreateEGLWindowSurface(surface, windowHandle)) {
+            state->DestroySurface(dpy, surface);
             state->SetError(EGL_BAD_NATIVE_WINDOW);
             return EGL_NO_SURFACE;
         }
 
-        return state->CreatePlatformWindowSurface(dpy, config, native_window, attrib_list);
+        return surface;
+    }
+
+    EGLBoolean ResizePlatformWindowSurface(EGLDisplay dpy, EGLSurface surface, EGLint width, EGLint height) {
+        auto* state = GetState();
+        if (!state) {
+            return EGL_FALSE;
+        }
+        if (!state->ResizeSurface(dpy, surface, width, height)) {
+            return EGL_FALSE;
+        }
+        auto* backendObject = GetBackendObject(state);
+        if (!backendObject) {
+            MGLOG_E("activeBackendObject not initialized!");
+            return EGL_FALSE;
+        }
+        width = std::max<EGLint>(width, 1);
+        height = std::max<EGLint>(height, 1);
+        if (!backendObject->ResizeEGLWindowSurface(surface, static_cast<Uint32>(width), static_cast<Uint32>(height))) {
+            state->SetError(EGL_BAD_NATIVE_WINDOW);
+            return EGL_FALSE;
+        }
+        return EGL_TRUE;
     }
 
     EGLSurface CreatePlatformPixmapSurface(EGLDisplay dpy, EGLConfig config, void* native_pixmap,
@@ -586,6 +759,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
         if (!name) {
             return nullptr;
         }
+        MobileGL::EnsureInitialized();
 
         MGLOG_D("eglGetProcAddress(%s)", name);
         void* proc = MG_Impl::GetProcAddress(name);

MobileGL/MG_Impl/EGLImpl/EGLImpl.h

@@ -57,6 +57,7 @@ namespace MobileGL::MG_Impl::EGLImpl {
     EGLDisplay GetPlatformDisplay(EGLenum platform, void* native_display, const EGLAttrib* attrib_list);
     EGLSurface CreatePlatformWindowSurface(EGLDisplay dpy, EGLConfig config, void* native_window,
                                            const EGLAttrib* attrib_list);
+    EGLBoolean ResizePlatformWindowSurface(EGLDisplay dpy, EGLSurface surface, EGLint width, EGLint height);
     EGLSurface CreatePlatformPixmapSurface(EGLDisplay dpy, EGLConfig config, void* native_pixmap,
                                            const EGLAttrib* attrib_list);
     EGLBoolean WaitSync(EGLDisplay dpy, EGLSync sync, EGLint flags);

MobileGL/MG_Impl/EGLImpl/Exporting/Definitions.cpp

@@ -235,6 +235,14 @@ MOBILEGL_EGL_API EGLDisplay eglGetPlatformDisplay(EGLenum platform, void* native
     return MobileGL::MG_Impl::EGLImpl::GetPlatformDisplay(platform, native_display, attrib_list);
 }
 
+MOBILEGL_EGL_API EGLDisplay eglGetPlatformDisplayEXT(EGLenum platform, void* native_display,
+                                                     const EGLint* attrib_list) {
+    MGLOG_D("eglGetPlatformDisplayEXT(platform=%u, native_display=%p, attrib_list=%p)", platform, native_display,
+            attrib_list);
+    return MobileGL::MG_Impl::EGLImpl::GetPlatformDisplay(
+        platform, native_display, reinterpret_cast<const EGLAttrib*>(attrib_list));
+}
+
 MOBILEGL_EGL_API EGLSurface eglCreatePlatformWindowSurface(EGLDisplay dpy, EGLConfig config, void* native_window,
                                                            const EGLAttrib* attrib_list) {
     MGLOG_D("eglCreatePlatformWindowSurface(dpy=%p, config=%p, native_window=%p, attrib_list=%p)", dpy, config,

MobileGL/MG_Impl/GLImpl/Buffer/GL_Buffer.h

@@ -12,19 +12,42 @@
 namespace MobileGL::MG_Impl::GLImpl {
     /* @INSERTION_POINT:FUNCTION_DECLARATION@ */
     void GetBufferParameteriv(GLenum target, GLenum pname, GLint* params);
+    void GetBufferParameteri64v(GLenum target, GLenum pname, GLint64* params);
+    void GetBufferPointerv(GLenum target, GLenum pname, void** params);
     GLboolean IsBuffer(GLuint buffer);
     void DeleteBuffers(GLsizei n, const GLuint* buffers);
     void FlushMappedBufferRange(GLenum target, GLintptr offset, GLsizeiptr length);
     GLboolean UnmapBuffer(GLenum target);
     void* MapBufferRange(GLenum target, GLintptr offset, GLsizeiptr length, GLbitfield access);
     void* MapBuffer(GLenum target, GLenum access);
+    void BufferStorage(GLenum target, GLsizeiptr size, const void* data, GLbitfield flags);
+    void CreateBuffers(GLsizei n, GLuint* buffers);
+    void NamedBufferStorage(GLuint buffer, GLsizeiptr size, const void* data, GLbitfield flags);
+    void NamedBufferData(GLuint buffer, GLsizeiptr size, const void* data, GLenum usage);
+    void NamedBufferSubData(GLuint buffer, GLintptr offset, GLsizeiptr size, const void* data);
+    void CopyNamedBufferSubData(GLuint readBuffer, GLuint writeBuffer, GLintptr readOffset, GLintptr writeOffset,
+                                GLsizeiptr size);
+    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);
+    void* MapNamedBuffer(GLuint buffer, GLenum access);
+    void* MapNamedBufferRange(GLuint buffer, GLintptr offset, GLsizeiptr length, GLbitfield access);
+    GLboolean UnmapNamedBuffer(GLuint buffer);
+    void FlushMappedNamedBufferRange(GLuint buffer, GLintptr offset, GLsizeiptr length);
+    void GetNamedBufferParameteriv(GLuint buffer, GLenum pname, GLint* params);
+    void GetNamedBufferParameteri64v(GLuint buffer, GLenum pname, GLint64* params);
+    void GetNamedBufferPointerv(GLuint buffer, GLenum pname, void** params);
     void CopyBufferSubData(GLenum readTarget, GLenum writeTarget, GLintptr readOffset, GLintptr writeOffset,
                            GLsizeiptr size);
     void BufferSubData(GLenum target, GLintptr offset, GLsizeiptr size, const void* data);
+    void GetBufferSubData(GLenum target, 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);
     void BindBufferBase(GLenum target, GLuint index, GLuint buffer);
     void BindBufferRange(GLenum target, GLuint index, GLuint buffer, GLintptr offset, GLsizeiptr size);
+    void BindBuffersBase(GLenum target, GLuint first, GLsizei count, const GLuint* buffers);
+    void BindBuffersRange(GLenum target, GLuint first, GLsizei count, const GLuint* buffers, const GLintptr* offsets,
+                          const GLsizeiptr* sizes);
 
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Buffer/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>
@@ -52,6 +53,26 @@ 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)));
+        }
+
+        if (index < pointCount) {
+            return true;
+        }
+
+        MG_State::pGLContext->RecordError(
+            ErrorCode::InvalidValue,
+            MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl/BufferImpl", "ValidateBufferBindingPointIndex",
+                                         std::format("Binding point index {} is out of range for target {}.", index,
+                                                     MG_Util::ConvertBufferTargetToString(target))));
+        return false;
+    }
+
     Bool ValidateBufferName(Uint index, Bool allowZero) {
         if (index == 0) {
             if (allowZero) return true;

MobileGL/MG_Impl/GLImpl/Buffer/Validators.h

@@ -16,4 +16,5 @@ namespace MobileGL::MG_Impl::GLImpl::BufferImpl {
     Bool ValidateBufferUsage(BufferUsage usage);
     Bool ValidateBufferMappingAccess(Flags<BufferMappingAccessBit> accessBits);
     Bool ValidateBufferBindingPointTarget(BufferTarget target);
+    Bool ValidateBufferBindingPointIndex(BufferTarget target, Uint index);
 } // namespace MobileGL::MG_Impl::GLImpl::BufferImpl

MobileGL/MG_Impl/GLImpl/Drawing/GL_Drawing.cpp

@@ -9,9 +9,75 @@
 #include "GL_Drawing.h"
 #include <Config.h>
 #include <MG_State/GLState/Core.h>
+#include <MG_State/EGLState/Core.h>
 #include <MG_Backend/BackendObjects.h>
 
 namespace MobileGL::MG_Impl::GLImpl {
+    static Bool ValidateCurrentProgramForExecution(const char* functionName) {
+        const auto& currentProgram = MG_State::pGLContext->GetCurrentProgram();
+        if (!currentProgram) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName, "There is no current program object."));
+            return false;
+        }
+
+        if (!currentProgram->GetLinkStatus()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                             "The current program object is not linked."));
+            return false;
+        }
+
+        return true;
+    }
+
+    static Bool ValidateCurrentProgramForCompute(const char* functionName) {
+        if (!ValidateCurrentProgramForExecution(functionName)) return false;
+
+        const auto& currentProgram = MG_State::pGLContext->GetCurrentProgram();
+        if (currentProgram->GetShaderIndexByStage(ShaderStage::Compute) < 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                             "The current program object has no compute shader stage."));
+            return false;
+        }
+
+        return true;
+    }
+
+    static Bool ValidatePrimitiveModeForBackend(const char* functionName, GLenum mode) {
+        const auto& activeBackendObject = MG_Backend::pActiveBackendObject;
+        if (!activeBackendObject) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName, "No active backend object."));
+            return false;
+        }
+
+        if (activeBackendObject->GetBackendType() == BackendType::DirectVulkan && mode == GL_LINE_LOOP) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl", functionName,
+                    "Primitive mode GL_LINE_LOOP is not supported by the DirectVulkan backend."));
+            return false;
+        }
+
+        const auto& vao = MG_State::pGLContext->GetBoundVertexArray();
+        if (vao && vao->GetExternalIndex() == 0 && !MG_State::IsRelaxedSemanticsActive()) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                             "Default vertex array object cannot be used for drawing in core profile."));
+            return false;
+        }
+
+        return true;
+    }
+
     void Clear_Backend(GLbitfield mask) {
 #ifdef TRACY_ENABLE
         ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
@@ -50,6 +116,13 @@ namespace MobileGL::MG_Impl::GLImpl {
         MG_Backend::gBackendFunctionsTable.GL.DrawArrays(mode, first, count);
     }
 
+    void MultiDrawArrays_Backend(GLenum mode, const GLint* first, const GLsizei* count, GLsizei drawcount) {
+#ifdef TRACY_ENABLE
+        ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+        MG_Backend::gBackendFunctionsTable.GL.MultiDrawArrays(mode, first, count, drawcount);
+    }
+
     void DrawElementsBaseVertex_Backend(GLenum mode, GLsizei count, GLenum type, const void* indices,
                                         GLint basevertex) {
 #ifdef TRACY_ENABLE
@@ -73,6 +146,24 @@ namespace MobileGL::MG_Impl::GLImpl {
         MG_Backend::gBackendFunctionsTable.GL.MultiDrawArraysIndirect(mode, indirect, drawcount, stride);
     }
 
+    void MultiDrawElementsIndirectCount_Backend(GLenum mode, GLenum type, const void* indirect, GLintptr drawcount,
+                                                GLsizei maxdrawcount, GLsizei stride) {
+#ifdef TRACY_ENABLE
+        ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+        MG_Backend::gBackendFunctionsTable.GL.MultiDrawElementsIndirectCount(mode, type, indirect, drawcount,
+                                                                             maxdrawcount, stride);
+    }
+
+    void MultiDrawArraysIndirectCount_Backend(GLenum mode, const void* indirect, GLintptr drawcount,
+                                              GLsizei maxdrawcount, GLsizei stride) {
+#ifdef TRACY_ENABLE
+        ZoneScopedC(TRACY_ZONECOLOR_BACKEND);
+#endif
+        MG_Backend::gBackendFunctionsTable.GL.MultiDrawArraysIndirectCount(mode, indirect, drawcount, maxdrawcount,
+                                                                           stride);
+    }
+
     void DrawRangeElementsBaseVertex_Backend(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type,
                                              const void* indices, GLint basevertex) {
 #ifdef TRACY_ENABLE
@@ -156,75 +247,193 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     /* @INSERTION_POINT:FUNCTION_IMPLEMENTATION@ */
+    void DispatchCompute(GLuint numGroupsX, GLuint numGroupsY, GLuint numGroupsZ) {
+        auto dispatchCompute = MG_Backend::gBackendFunctionsTable.GL.DispatchCompute;
+        if (!dispatchCompute) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "Backend does not support compute dispatch."));
+            return;
+        }
+        if (!ValidateCurrentProgramForCompute(__func__)) return;
+        dispatchCompute(numGroupsX, numGroupsY, numGroupsZ);
+    }
+
+    void DispatchComputeIndirect(GLintptr indirect) {
+        auto dispatchComputeIndirect = MG_Backend::gBackendFunctionsTable.GL.DispatchComputeIndirect;
+        if (!dispatchComputeIndirect) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Backend does not support indirect compute dispatch."));
+            return;
+        }
+        if (!ValidateCurrentProgramForCompute(__func__)) return;
+        dispatchComputeIndirect(indirect);
+    }
+
+    void MemoryBarrier(GLbitfield barriers) {
+        auto memoryBarrier = MG_Backend::gBackendFunctionsTable.GL.MemoryBarrier;
+        if (!memoryBarrier) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "Backend does not support memory barriers."));
+            return;
+        }
+        memoryBarrier(barriers);
+    }
+
+    void MemoryBarrierByRegion(GLbitfield barriers) {
+        auto memoryBarrierByRegion = MG_Backend::gBackendFunctionsTable.GL.MemoryBarrierByRegion;
+        if (!memoryBarrierByRegion) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Backend does not support regional memory barriers."));
+            return;
+        }
+        memoryBarrierByRegion(barriers);
+    }
+
     void MultiDrawElementsIndirect(GLenum mode, GLenum type, const void* indirect, GLsizei drawcount, GLsizei stride) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         MultiDrawElementsIndirect_Backend(mode, type, indirect, drawcount, stride);
     }
 
     void MultiDrawArraysIndirect(GLenum mode, const void* indirect, GLsizei drawcount, GLsizei stride) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         MultiDrawArraysIndirect_Backend(mode, indirect, drawcount, stride);
     }
 
+    void MultiDrawElementsIndirectCount(GLenum mode, GLenum type, const void* indirect, GLintptr drawcount,
+                                        GLsizei maxdrawcount, GLsizei stride) {
+        auto multiDrawElementsIndirectCount = MG_Backend::gBackendFunctionsTable.GL.MultiDrawElementsIndirectCount;
+        if (!multiDrawElementsIndirectCount) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Backend does not support indirect-parameter indexed draws."));
+            return;
+        }
+        MultiDrawElementsIndirectCount_Backend(mode, type, indirect, drawcount, maxdrawcount, stride);
+    }
+
+    void MultiDrawArraysIndirectCount(GLenum mode, const void* indirect, GLintptr drawcount,
+                                      GLsizei maxdrawcount, GLsizei stride) {
+        auto multiDrawArraysIndirectCount = MG_Backend::gBackendFunctionsTable.GL.MultiDrawArraysIndirectCount;
+        if (!multiDrawArraysIndirectCount) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Backend does not support indirect-parameter array draws."));
+            return;
+        }
+        MultiDrawArraysIndirectCount_Backend(mode, indirect, drawcount, maxdrawcount, stride);
+    }
+
     void DrawRangeElementsBaseVertex(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type,
                                      const void* indices, GLint basevertex) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawRangeElementsBaseVertex_Backend(mode, start, end, count, type, indices, basevertex);
     }
 
     void DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type, const void* indices) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawRangeElements_Backend(mode, start, end, count, type, indices);
     }
 
     void DrawElementsInstancedBaseVertexBaseInstance(GLenum mode, GLsizei count, GLenum type, const void* indices,
                                                      GLsizei instancecount, GLint basevertex, GLuint baseinstance) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawElementsInstancedBaseVertexBaseInstance_Backend(mode, count, type, indices, instancecount, basevertex,
                                                             baseinstance);
     }
 
     void DrawElementsInstancedBaseVertex(GLenum mode, GLsizei count, GLenum type, const void* indices,
                                          GLsizei instancecount, GLint basevertex) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawElementsInstancedBaseVertex_Backend(mode, count, type, indices, instancecount, basevertex);
     }
 
     void DrawElementsInstancedBaseInstance(GLenum mode, GLsizei count, GLenum type, const void* indices,
                                            GLsizei instancecount, GLuint baseinstance) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawElementsInstancedBaseInstance_Backend(mode, count, type, indices, instancecount, baseinstance);
     }
 
     void DrawElementsInstanced(GLenum mode, GLsizei count, GLenum type, const void* indices, GLsizei instancecount) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawElementsInstanced_Backend(mode, count, type, indices, instancecount);
     }
 
     void DrawElementsIndirect(GLenum mode, GLenum type, const void* indirect) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawElementsIndirect_Backend(mode, type, indirect);
     }
 
     void DrawArraysInstancedBaseInstance(GLenum mode, GLint first, GLsizei count, GLsizei instancecount,
                                          GLuint baseinstance) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawArraysInstancedBaseInstance_Backend(mode, first, count, instancecount, baseinstance);
     }
 
     void DrawArraysInstanced(GLenum mode, GLint first, GLsizei count, GLsizei instancecount) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawArraysInstanced_Backend(mode, first, count, instancecount);
     }
 
     void DrawArraysIndirect(GLenum mode, const void* indirect) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawArraysIndirect_Backend(mode, indirect);
     }
 
     void DrawElementsBaseVertex(GLenum mode, GLsizei count, GLenum type, const void* indices, GLint basevertex) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawElementsBaseVertex_Backend(mode, count, type, indices, basevertex);
     }
 
     void DrawArrays(GLenum mode, GLint first, GLsizei count) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawArrays_Backend(mode, first, count);
     }
 
+    void MultiDrawArrays(GLenum mode, const GLint* first, const GLsizei* count, GLsizei drawcount) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
+        if (drawcount < 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "drawcount must be non-negative."));
+            return;
+        }
+        MultiDrawArrays_Backend(mode, first, count, drawcount);
+    }
+
     void MultiDrawElements(GLenum mode, const GLsizei* count, GLenum type, const void* const* indices,
                            GLsizei drawcount) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         MultiDrawElements_Backend(mode, count, type, indices, drawcount);
     }
 
     void MultiDrawElementsBaseVertex(GLenum mode, const GLsizei* count, GLenum type, const void* const* indices,
                                      GLsizei drawcount, const GLint* basevertex) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         MultiDrawElementsBaseVertex_Backend(mode, count, type, indices, drawcount, basevertex);
     }
 
@@ -233,6 +442,8 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void DrawElements(GLenum mode, GLsizei count, GLenum type, const void* indices) {
+        if (!ValidateCurrentProgramForExecution(__func__)) return;
+        if (!ValidatePrimitiveModeForBackend(__func__, mode)) return;
         DrawElements_Backend(mode, count, type, indices);
     }
 

MobileGL/MG_Impl/GLImpl/Drawing/GL_Drawing.h

@@ -11,8 +11,16 @@
 
 namespace MobileGL::MG_Impl::GLImpl {
     /* @INSERTION_POINT:FUNCTION_DECLARATION@ */
+    void DispatchCompute(GLuint numGroupsX, GLuint numGroupsY, GLuint numGroupsZ);
+    void DispatchComputeIndirect(GLintptr indirect);
+    void MemoryBarrier(GLbitfield barriers);
+    void MemoryBarrierByRegion(GLbitfield barriers);
     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,
+                                        GLsizei maxdrawcount, 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);
@@ -30,6 +38,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     void DrawArraysIndirect(GLenum mode, const void* indirect);
     void DrawElementsBaseVertex(GLenum mode, GLsizei count, GLenum type, const void* indices, GLint basevertex);
     void DrawArrays(GLenum mode, GLint first, GLsizei count);
+    void MultiDrawArrays(GLenum mode, const GLint* first, const GLsizei* count, GLsizei drawcount);
     void MultiDrawElements(GLenum mode, const GLsizei* count, GLenum type, const void* const* indices,
                            GLsizei drawcount);
     void MultiDrawElementsBaseVertex(GLenum mode, const GLsizei* count, GLenum type, const void* const* indices,

MobileGL/MG_Impl/GLImpl/Framebuffer/GL_Framebuffer.h

@@ -24,19 +24,43 @@ namespace MobileGL::MG_Impl::GLImpl {
     GLboolean IsRenderbuffer(GLuint renderbuffer);
     void GetRenderbufferParameteriv(GLenum target, GLenum pname, GLint* params);
     void GenRenderbuffers(GLsizei n, GLuint* renderbuffers);
+    void CreateRenderbuffers(GLsizei n, GLuint* renderbuffers);
+    void NamedRenderbufferStorage(GLuint renderbuffer, GLenum internalformat, GLsizei width, GLsizei height);
+    void NamedRenderbufferStorageMultisample(GLuint renderbuffer, GLsizei samples, GLenum internalformat,
+                                             GLsizei width, GLsizei height);
+    void GetNamedRenderbufferParameteriv(GLuint renderbuffer, GLenum pname, GLint* params);
     void FramebufferRenderbuffer(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer);
+    void NamedFramebufferRenderbuffer(GLuint framebuffer, GLenum attachment, GLenum renderbuffertarget,
+                                      GLuint renderbuffer);
     void DeleteRenderbuffers(GLsizei n, const GLuint* renderbuffers);
     void BindRenderbuffer(GLenum target, GLuint renderbuffer);
     void SampleMaski(GLuint maskNumber, GLbitfield mask);
     GLboolean IsFramebuffer(GLuint framebuffer);
     void GetFramebufferAttachmentParameteriv(GLenum target, GLenum attachment, GLenum pname, GLint* params);
     void GenFramebuffers(GLsizei n, GLuint* framebuffers);
+    void CreateFramebuffers(GLsizei n, GLuint* framebuffers);
     void FramebufferTextureLayer(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer);
     void FramebufferTexture3D(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level,
                               GLint zoffset);
     void FramebufferTexture2D(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
     void FramebufferTexture1D(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
     void FramebufferTexture(GLenum target, GLenum attachment, GLuint texture, GLint level);
+    void NamedFramebufferTexture(GLuint framebuffer, GLenum attachment, GLuint texture, GLint level);
+    void NamedFramebufferTexture1D(GLuint framebuffer, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
+    void NamedFramebufferTexture2D(GLuint framebuffer, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
+    void NamedFramebufferTexture3D(GLuint framebuffer, GLenum attachment, GLenum textarget, GLuint texture, GLint level,
+                                   GLint zoffset);
+    void NamedFramebufferTextureLayer(GLuint framebuffer, GLenum attachment, GLuint texture, GLint level, GLint layer);
+    void NamedFramebufferDrawBuffer(GLuint framebuffer, GLenum buf);
+    void NamedFramebufferDrawBuffers(GLuint framebuffer, GLsizei n, const GLenum* bufs);
+    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);
+    GLenum CheckNamedFramebufferStatus(GLuint framebuffer, GLenum target);
+    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,
+                              GLenum filter);
     void DrawBuffer(GLenum buf);
     void DrawBuffers(GLsizei n, const GLenum* bufs);
     void ReadBuffer(GLenum src);
@@ -54,6 +78,6 @@ 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

@@ -76,7 +76,13 @@ namespace MobileGL::MG_Impl::GLImpl::FramebufferImpl {
     }
 
     Bool ValidateRenderbufferName(Uint index, Bool allowZero) {
-        if (index == 0 && !allowZero) {
+        if (index == 0) {
+            // Zero is never a GenRenderbuffers name, so it must not reach the name-table lookup
+            // below: where it is allowed (glBindRenderbuffer / FramebufferRenderbuffer detach) it
+            // means "unbind", and looking it up would record a bogus INVALID_OPERATION - GL CTS's
+            // per-case state reset calls glBindRenderbuffer(GL_RENDERBUFFER, 0) after every case.
+            if (allowZero) return true;
+
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue,
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl/FramebufferImpl", "ValidateRenderbufferName",

MobileGL/MG_Impl/GLImpl/Getter/GL_Getter.h

@@ -13,6 +13,13 @@ namespace MobileGL::MG_Impl::GLImpl {
     /* @INSERTION_POINT:FUNCTION_DECLARATION@ */
     const GLubyte* GetString(GLenum name);
     const GLubyte* GetStringi(GLenum name, GLuint index);
+    void GetBooleanv(GLenum pname, GLboolean* params);
+    void GetFloatv(GLenum pname, GLfloat* params);
+    void GetDoublev(GLenum pname, GLdouble* params);
     void GetIntegerv(GLenum pname, GLint* params);
+    void GetInteger64v(GLenum pname, GLint64* params);
+    void GetIntegeri_v(GLenum target, GLuint index, GLint* data);
+    void GetInteger64i_v(GLenum target, GLuint index, GLint64* data);
     GLenum GetError();
+    GLenum GetGraphicsResetStatus();
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Program/GL_Program.h

@@ -22,6 +22,12 @@ namespace MobileGL::MG_Impl::GLImpl {
                          GLchar* name);
     void GetActiveUniform(GLuint program, GLuint index, GLsizei bufSize, GLsizei* length, GLint* size, GLenum* type,
                           GLchar* name);
+    void GetActiveUniformName(GLuint program, GLuint uniformIndex, GLsizei bufSize, GLsizei* length,
+                              GLchar* uniformName);
+    void GetUniformIndices(GLuint program, GLsizei uniformCount, const GLchar* const* uniformNames,
+                           GLuint* uniformIndices);
+    void GetActiveUniformsiv(GLuint program, GLsizei uniformCount, const GLuint* uniformIndices, GLenum pname,
+                             GLint* params);
     void GetAttachedShaders(GLuint program, GLsizei maxCount, GLsizei* count, GLuint* shaders);
     GLint GetAttribLocation(GLuint program, const GLchar* name);
     void GetProgramiv(GLuint program, GLenum pname, GLint* params);
@@ -32,6 +38,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     GLint GetUniformLocation(GLuint program, const GLchar* name);
     void GetUniformfv(GLuint program, GLint location, GLfloat* params);
     void GetUniformiv(GLuint program, GLint location, GLint* params);
+    void GetUniformuiv(GLuint program, GLint location, GLuint* params);
     GLboolean IsProgram(GLuint program);
     GLboolean IsShader(GLuint shader);
     void LinkProgram(GLuint program);
@@ -45,6 +52,10 @@ namespace MobileGL::MG_Impl::GLImpl {
     void Uniform2i(GLint location, GLint v0, GLint v1);
     void Uniform3i(GLint location, GLint v0, GLint v1, GLint v2);
     void Uniform4i(GLint location, GLint v0, GLint v1, GLint v2, GLint v3);
+    void Uniform1ui(GLint location, GLuint v0);
+    void Uniform2ui(GLint location, GLuint v0, GLuint v1);
+    void Uniform3ui(GLint location, GLuint v0, GLuint v1, GLuint v2);
+    void Uniform4ui(GLint location, GLuint v0, GLuint v1, GLuint v2, GLuint v3);
     void Uniform1fv(GLint location, GLsizei count, const GLfloat* value);
     void Uniform2fv(GLint location, GLsizei count, const GLfloat* value);
     void Uniform3fv(GLint location, GLsizei count, const GLfloat* value);
@@ -53,15 +64,78 @@ namespace MobileGL::MG_Impl::GLImpl {
     void Uniform2iv(GLint location, GLsizei count, const GLint* value);
     void Uniform3iv(GLint location, GLsizei count, const GLint* value);
     void Uniform4iv(GLint location, GLsizei count, const GLint* value);
+    void Uniform1uiv(GLint location, GLsizei count, const GLuint* value);
+    void Uniform2uiv(GLint location, GLsizei count, const GLuint* value);
+    void Uniform3uiv(GLint location, GLsizei count, const GLuint* value);
+    void Uniform4uiv(GLint location, GLsizei count, const GLuint* value);
     void UniformMatrix2fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
     void UniformMatrix3fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
     void UniformMatrix4fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
+    void ProgramUniform1f(GLuint program, GLint location, GLfloat v0);
+    void ProgramUniform2f(GLuint program, GLint location, GLfloat v0, GLfloat v1);
+    void ProgramUniform3f(GLuint program, GLint location, GLfloat v0, GLfloat v1, GLfloat v2);
+    void ProgramUniform4f(GLuint program, GLint location, GLfloat v0, GLfloat v1, GLfloat v2, GLfloat v3);
+    void ProgramUniform1i(GLuint program, GLint location, GLint v0);
+    void ProgramUniform2i(GLuint program, GLint location, GLint v0, GLint v1);
+    void ProgramUniform3i(GLuint program, GLint location, GLint v0, GLint v1, GLint v2);
+    void ProgramUniform4i(GLuint program, GLint location, GLint v0, GLint v1, GLint v2, GLint v3);
+    void ProgramUniform1ui(GLuint program, GLint location, GLuint v0);
+    void ProgramUniform2ui(GLuint program, GLint location, GLuint v0, GLuint v1);
+    void ProgramUniform3ui(GLuint program, GLint location, GLuint v0, GLuint v1, GLuint v2);
+    void ProgramUniform4ui(GLuint program, GLint location, GLuint v0, GLuint v1, GLuint v2, GLuint v3);
+    void ProgramUniform1fv(GLuint program, GLint location, GLsizei count, const GLfloat* value);
+    void ProgramUniform2fv(GLuint program, GLint location, GLsizei count, const GLfloat* value);
+    void ProgramUniform3fv(GLuint program, GLint location, GLsizei count, const GLfloat* value);
+    void ProgramUniform4fv(GLuint program, GLint location, GLsizei count, const GLfloat* value);
+    void ProgramUniform1iv(GLuint program, GLint location, GLsizei count, const GLint* value);
+    void ProgramUniform2iv(GLuint program, GLint location, GLsizei count, const GLint* value);
+    void ProgramUniform3iv(GLuint program, GLint location, GLsizei count, const GLint* value);
+    void ProgramUniform4iv(GLuint program, GLint location, GLsizei count, const GLint* value);
+    void ProgramUniform1uiv(GLuint program, GLint location, GLsizei count, const GLuint* value);
+    void ProgramUniform2uiv(GLuint program, GLint location, GLsizei count, const GLuint* value);
+    void ProgramUniform3uiv(GLuint program, GLint location, GLsizei count, const GLuint* value);
+    void ProgramUniform4uiv(GLuint program, GLint location, GLsizei count, const GLuint* value);
+    void ProgramUniformMatrix2fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                 const GLfloat* value);
+    void ProgramUniformMatrix3fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                 const GLfloat* value);
+    void ProgramUniformMatrix4fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                 const GLfloat* value);
+    void UniformMatrix2x3fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
+    void UniformMatrix3x2fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
+    void UniformMatrix2x4fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
+    void UniformMatrix4x2fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
+    void UniformMatrix3x4fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
+    void UniformMatrix4x3fv(GLint location, GLsizei count, GLboolean transpose, const GLfloat* value);
+    void ProgramUniformMatrix2x3fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                   const GLfloat* value);
+    void ProgramUniformMatrix3x2fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                   const GLfloat* value);
+    void ProgramUniformMatrix2x4fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                   const GLfloat* value);
+    void ProgramUniformMatrix4x2fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                   const GLfloat* value);
+    void ProgramUniformMatrix3x4fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                   const GLfloat* value);
+    void ProgramUniformMatrix4x3fv(GLuint program, GLint location, GLsizei count, GLboolean transpose,
+                                   const GLfloat* value);
     GLuint GetUniformBlockIndex(GLuint program, const GLchar* uniformBlockName);
     void UniformBlockBinding(GLuint program, GLuint uniformBlockIndex, GLuint uniformBlockBinding);
     void GetActiveUniformBlockiv(GLuint program, GLuint uniformBlockIndex, GLenum pname, GLint* params);
     void GetActiveUniformBlockName(GLuint program, GLuint uniformBlockIndex, GLsizei bufSize, GLsizei* length,
                                    GLchar* uniformBlockName);
     void BindFragDataLocation(GLuint program, GLuint colorNumber, const char* name);
+    void BindFragDataLocationIndexed(GLuint program, GLuint colorNumber, GLuint index, const char* name);
     GLint GetFragDataLocation(GLuint program, const char* name);
+    GLint GetFragDataIndex(GLuint program, const char* name);
+    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 ValidateProgram(GLuint program);
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Query/GL_Query.cpp

@@ -0,0 +1,362 @@
+// MobileGL - MobileGL/MG_Impl/GLImpl/Query/GL_Query.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_Query.h"
+#include <Config.h>
+#include <MG_Backend/BackendObjects.h>
+#include <MG_State/GLState/Core.h>
+#include <MG_State/GLState/ErrorState/ErrorInfo.h>
+
+namespace MobileGL::MG_Impl::GLImpl {
+    namespace {
+        // Frontend query object (GL_ARB_timer_query): wraps an optional backend
+        // timer-query handle. A null backend handle (backend has no timer-query
+        // support, timer queries are disabled by config, or the backend could
+        // not create a query at call time) keeps a graceful fallback: the query
+        // result is immediately available and reads as zero.
+        struct QueryObject {
+            GLuint id = 0;
+            GLenum target = 0; // 0 = gen'd but never used with BeginQuery/QueryCounter
+            MG_Backend::BackendQueryHandle backendHandle = nullptr;
+            Bool active = false;
+            Bool ended = false;
+            Bool resultCached = false;
+            Uint64 cachedResult = 0;
+        };
+
+        // Query calls may arrive from any thread (launchers migrate the context
+        // across JVM threads), so the live-object registry is mutex-guarded,
+        // like the sync-object registry in GL_Sync.cpp. Entries left at process
+        // shutdown are simply dropped; their backend handles die with the
+        // backend.
+        std::mutex g_queryObjectsMutex;
+        UnorderedMap<GLuint, QueryObject*> g_liveQueryObjects;
+        // Monotonically increasing id allocator; ids are valid query objects
+        // immediately after GenQueries.
+        GLuint g_nextQueryId = 1;
+        // Id of the query currently active on GL_TIME_ELAPSED (0 = none).
+        GLuint g_activeTimeElapsedQueryId = 0;
+
+        Bool TimerQueryDisabled() {
+            return MG_Config::Features.DisableTimerQuery;
+        }
+
+        void RecordQueryError(ErrorCode code, const char* function, const char* message) {
+            MG_State::pGLContext->RecordError(code,
+                                              MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", function, message));
+        }
+
+        // Callers must hold g_queryObjectsMutex.
+        QueryObject* FindQueryObjectLocked(GLuint id) {
+            const auto it = g_liveQueryObjects.find(id);
+            return it != g_liveQueryObjects.end() ? it->second : nullptr;
+        }
+
+        // Callers must hold g_queryObjectsMutex. Releases the backend handle
+        // (if any) and clears any cached result, so the object can be reused.
+        void ResetQueryObjectLocked(QueryObject* queryObject) {
+            if (queryObject->backendHandle) {
+                if (const auto deleteBackendQuery = MG_Backend::gBackendFunctionsTable.GL.DeleteBackendQuery) {
+                    deleteBackendQuery(queryObject->backendHandle);
+                }
+                queryObject->backendHandle = nullptr;
+            }
+            queryObject->active = false;
+            queryObject->ended = false;
+            queryObject->resultCached = false;
+            queryObject->cachedResult = 0;
+        }
+
+        // Callers must hold g_queryObjectsMutex.
+        void EndTimeElapsedQueryLocked(QueryObject* queryObject) {
+            const auto endTimeElapsedQuery = MG_Backend::gBackendFunctionsTable.GL.EndTimeElapsedQuery;
+            if (endTimeElapsedQuery && queryObject->backendHandle) {
+                endTimeElapsedQuery(queryObject->backendHandle);
+            }
+            queryObject->active = false;
+            queryObject->ended = true;
+            g_activeTimeElapsedQueryId = 0;
+        }
+
+        // Shared GetQueryObject* implementation. Returns false when an error
+        // was recorded and no value should be written back.
+        Bool GetQueryObjectValue(GLuint id, GLenum pname, const char* function, Uint64& outValue) {
+            const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+            auto* queryObject = FindQueryObjectLocked(id);
+            if (!queryObject) {
+                RecordQueryError(ErrorCode::InvalidOperation, function, "Query object does not exist.");
+                return false;
+            }
+            if (queryObject->active) {
+                RecordQueryError(ErrorCode::InvalidOperation, function, "Query object is still active.");
+                return false;
+            }
+
+            switch (pname) {
+            case GL_QUERY_RESULT_AVAILABLE: {
+                if (queryObject->resultCached || !queryObject->backendHandle) {
+                    outValue = 1;
+                    return true;
+                }
+                const auto isQueryResultAvailable = MG_Backend::gBackendFunctionsTable.GL.IsQueryResultAvailable;
+                outValue = (!isQueryResultAvailable || isQueryResultAvailable(queryObject->backendHandle)) ? 1 : 0;
+                return true;
+            }
+            case GL_QUERY_RESULT: {
+                if (queryObject->resultCached) {
+                    outValue = queryObject->cachedResult;
+                    return true;
+                }
+                Uint64 result = 0;
+                if (queryObject->backendHandle) {
+                    const auto getQueryResult64 = MG_Backend::gBackendFunctionsTable.GL.GetQueryResult64;
+                    if (getQueryResult64 &&
+                        !getQueryResult64(queryObject->backendHandle, /*wait=*/true, &result)) {
+                        // The backend could not produce the result YET (e.g. a
+                        // Vulkan wait refusing to block on a not-yet-submitted
+                        // frame serial). Per the documented no-stall tradeoff
+                        // this call reads 0, but the value is NOT cached and
+                        // the backend handle is kept, so a later AVAILABLE
+                        // poll / RESULT read still produces the real value.
+                        outValue = 0;
+                        return true;
+                    }
+                    // Final value produced (or no GetQueryResult64 hook: the
+                    // query degrades to a zero result); the backend handle is
+                    // consumed and the value cached for later reads.
+                    if (const auto deleteBackendQuery = MG_Backend::gBackendFunctionsTable.GL.DeleteBackendQuery) {
+                        deleteBackendQuery(queryObject->backendHandle);
+                    }
+                    queryObject->backendHandle = nullptr;
+                }
+                queryObject->cachedResult = result;
+                queryObject->resultCached = true;
+                outValue = result;
+                return true;
+            }
+            default:
+                RecordQueryError(ErrorCode::InvalidEnum, function, "Unsupported query object parameter.");
+                return false;
+            }
+        }
+    } // namespace
+
+    void GenQueries(GLsizei n, GLuint* ids) {
+        if (n < 0) {
+            RecordQueryError(ErrorCode::InvalidValue, __FUNCTION__, "n cannot be negative.");
+            return;
+        }
+        if (!ids) {
+            return;
+        }
+        const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+        for (GLsizei i = 0; i < n; ++i) {
+            const GLuint id = g_nextQueryId++;
+            auto* queryObject = new QueryObject;
+            queryObject->id = id;
+            g_liveQueryObjects[id] = queryObject;
+            ids[i] = id;
+        }
+    }
+
+    void DeleteQueries(GLsizei n, const GLuint* ids) {
+        if (n < 0) {
+            RecordQueryError(ErrorCode::InvalidValue, __FUNCTION__, "n cannot be negative.");
+            return;
+        }
+        if (!ids) {
+            return;
+        }
+        const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+        for (GLsizei i = 0; i < n; ++i) {
+            const auto it = g_liveQueryObjects.find(ids[i]);
+            if (it == g_liveQueryObjects.end()) {
+                continue; // unknown ids are silently ignored
+            }
+            QueryObject* queryObject = it->second;
+            if (queryObject->active) {
+                EndTimeElapsedQueryLocked(queryObject); // implicitly end before deletion
+            }
+            if (queryObject->backendHandle) {
+                if (const auto deleteBackendQuery = MG_Backend::gBackendFunctionsTable.GL.DeleteBackendQuery) {
+                    deleteBackendQuery(queryObject->backendHandle);
+                }
+                queryObject->backendHandle = nullptr;
+            }
+            g_liveQueryObjects.erase(it);
+            delete queryObject;
+        }
+    }
+
+    GLboolean IsQuery(GLuint id) {
+        if (id == 0) {
+            return GL_FALSE;
+        }
+        const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+        // Gen'd ids count as query objects here: the registry creates live
+        // objects at GenQueries time.
+        return FindQueryObjectLocked(id) != nullptr ? GL_TRUE : GL_FALSE;
+    }
+
+    void BeginQuery(GLenum target, GLuint id) {
+        if (target != GL_TIME_ELAPSED) {
+            // Only GL_TIME_ELAPSED timer queries are implemented (occlusion and
+            // primitive queries remain stubs); GL_TIMESTAMP is not a valid
+            // BeginQuery target either.
+            RecordQueryError(ErrorCode::InvalidEnum, __FUNCTION__, "Query target is not supported.");
+            return;
+        }
+        if (id == 0) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__, "Query id 0 cannot be used.");
+            return;
+        }
+        const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+        auto* queryObject = FindQueryObjectLocked(id);
+        if (!queryObject) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__, "Query object does not exist.");
+            return;
+        }
+        if (g_activeTimeElapsedQueryId != 0) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__,
+                             "A query is already active on GL_TIME_ELAPSED.");
+            return;
+        }
+        if (queryObject->active) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__, "Query object is already active.");
+            return;
+        }
+        if (queryObject->target != 0 && queryObject->target != target) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__,
+                             "Query object was already used with a different target.");
+            return;
+        }
+
+        ResetQueryObjectLocked(queryObject); // discard any previous result
+        queryObject->target = target;
+        queryObject->active = true;
+        const auto beginTimeElapsedQuery = MG_Backend::gBackendFunctionsTable.GL.BeginTimeElapsedQuery;
+        queryObject->backendHandle =
+            (!TimerQueryDisabled() && beginTimeElapsedQuery) ? beginTimeElapsedQuery() : nullptr;
+        g_activeTimeElapsedQueryId = id;
+    }
+
+    void EndQuery(GLenum target) {
+        if (target != GL_TIME_ELAPSED) {
+            RecordQueryError(ErrorCode::InvalidEnum, __FUNCTION__, "Query target is not supported.");
+            return;
+        }
+        const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+        if (g_activeTimeElapsedQueryId == 0) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__, "No query is active on GL_TIME_ELAPSED.");
+            return;
+        }
+        auto* queryObject = FindQueryObjectLocked(g_activeTimeElapsedQueryId);
+        if (!queryObject) {
+            g_activeTimeElapsedQueryId = 0; // should not happen; keep state consistent
+            return;
+        }
+        EndTimeElapsedQueryLocked(queryObject);
+    }
+
+    void QueryCounter(GLuint id, GLenum target) {
+        if (target != GL_TIMESTAMP) {
+            RecordQueryError(ErrorCode::InvalidEnum, __FUNCTION__, "QueryCounter target must be GL_TIMESTAMP.");
+            return;
+        }
+        if (id == 0) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__, "Query id 0 cannot be used.");
+            return;
+        }
+        const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+        auto* queryObject = FindQueryObjectLocked(id);
+        if (!queryObject) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__, "Query object does not exist.");
+            return;
+        }
+        if (queryObject->active) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__, "Query object is currently active.");
+            return;
+        }
+        if (queryObject->target != 0 && queryObject->target != target) {
+            RecordQueryError(ErrorCode::InvalidOperation, __FUNCTION__,
+                             "Query object was already used with a different target.");
+            return;
+        }
+
+        ResetQueryObjectLocked(queryObject); // discard any previous result
+        queryObject->target = target;
+        const auto queryCounterTimestamp = MG_Backend::gBackendFunctionsTable.GL.QueryCounterTimestamp;
+        queryObject->backendHandle =
+            (!TimerQueryDisabled() && queryCounterTimestamp) ? queryCounterTimestamp() : nullptr;
+        queryObject->ended = true;
+    }
+
+    void GetQueryiv(GLenum target, GLenum pname, GLint* params) {
+        if (!params) {
+            return;
+        }
+        switch (pname) {
+        case GL_CURRENT_QUERY: {
+            const std::lock_guard<std::mutex> lock(g_queryObjectsMutex);
+            // Only GL_TIME_ELAPSED queries can be active; GL_TIMESTAMP queries
+            // never are, and other targets remain unimplemented.
+            *params = target == GL_TIME_ELAPSED ? static_cast<GLint>(g_activeTimeElapsedQueryId) : 0;
+            return;
+        }
+        case GL_QUERY_COUNTER_BITS: {
+            // 64 bits are advertised only while the live backend can actually
+            // time: IsTimerQuerySupported is the dynamic truth (extension /
+            // entry points / timestamp valid bits at call time, not at table
+            // init), and the MOBILEGL_DISABLE_TIMERQUERY kill switch always
+            // wins. Non-timer targets remain unimplemented and report 0.
+            const Bool timerTarget = target == GL_TIME_ELAPSED || target == GL_TIMESTAMP;
+            const auto isTimerQuerySupported = MG_Backend::gBackendFunctionsTable.GL.IsTimerQuerySupported;
+            const Bool supported =
+                timerTarget && !TimerQueryDisabled() && isTimerQuerySupported && isTimerQuerySupported();
+            *params = supported ? 64 : 0;
+            return;
+        }
+        default:
+            RecordQueryError(ErrorCode::InvalidEnum, __FUNCTION__, "Unsupported query parameter.");
+            return;
+        }
+    }
+
+    void GetQueryObjectiv(GLuint id, GLenum pname, GLint* params) {
+        Uint64 value = 0;
+        if (!GetQueryObjectValue(id, pname, __FUNCTION__, value) || !params) {
+            return;
+        }
+        constexpr Uint64 kMaxInt = static_cast<Uint64>(INT_MAX);
+        *params = value > kMaxInt ? INT_MAX : static_cast<GLint>(value);
+    }
+
+    void GetQueryObjectuiv(GLuint id, GLenum pname, GLuint* params) {
+        Uint64 value = 0;
+        if (!GetQueryObjectValue(id, pname, __FUNCTION__, value) || !params) {
+            return;
+        }
+        *params = static_cast<GLuint>(value & 0xFFFFFFFFull);
+    }
+
+    void GetQueryObjecti64v(GLuint id, GLenum pname, GLint64* params) {
+        Uint64 value = 0;
+        if (!GetQueryObjectValue(id, pname, __FUNCTION__, value) || !params) {
+            return;
+        }
+        *params = static_cast<GLint64>(value);
+    }
+
+    void GetQueryObjectui64v(GLuint id, GLenum pname, GLuint64* params) {
+        Uint64 value = 0;
+        if (!GetQueryObjectValue(id, pname, __FUNCTION__, value) || !params) {
+            return;
+        }
+        *params = static_cast<GLuint64>(value);
+    }
+} // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Query/GL_Query.h

@@ -0,0 +1,24 @@
+// MobileGL - MobileGL/MG_Impl/GLImpl/Query/GL_Query.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 {
+    void GenQueries(GLsizei n, GLuint* ids);
+    void DeleteQueries(GLsizei n, const GLuint* ids);
+    GLboolean IsQuery(GLuint id);
+    void BeginQuery(GLenum target, GLuint id);
+    void EndQuery(GLenum target);
+    void GetQueryiv(GLenum target, GLenum pname, GLint* params);
+    void GetQueryObjectiv(GLuint id, GLenum pname, GLint* params);
+    void GetQueryObjectuiv(GLuint id, GLenum pname, GLuint* params);
+    void GetQueryObjecti64v(GLuint id, GLenum pname, GLint64* params);
+    void GetQueryObjectui64v(GLuint id, GLenum pname, GLuint64* params);
+    void QueryCounter(GLuint id, GLenum target);
+} // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/RenderState/GL_RenderState.cpp

@@ -7,6 +7,7 @@
 // End of Source File Header
 
 #include "GL_RenderState.h"
+#include <cmath>
 #include <MG_State/GLState/Core.h>
 #include <MG_Util/Converters/GLToStr/GLEnumConverter.h>
 #include <MG_Util/Converters/GLToMG/RenderStateEnumConverter.h>
@@ -14,6 +15,82 @@
 #include <MG_Util/Converters/MGToStr/RenderStateEnumConverter.h>
 
 namespace MobileGL::MG_Impl::GLImpl {
+    static Float ClampUnitFloat(GLfloat value) {
+        return std::clamp(static_cast<Float>(value), 0.0f, 1.0f);
+    }
+
+    static Bool ValidateIndexedBlendCapability(GLenum target, GLuint index, const char* functionName) {
+        if (target != GL_BLEND) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                             "Only GL_BLEND is supported for indexed capability state."));
+            return false;
+        }
+
+        if (index >= MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl", functionName,
+                    "Buffer index " + std::to_string(index) + " is out of range. Max supported is " +
+                        std::to_string(MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS - 1) + "."));
+            return false;
+        }
+
+        return true;
+    }
+
+    static Bool TryConvertBlendEquation(GLenum mode, const char* functionName,
+                                        ::MobileGL::BlendEquation& outEquation) {
+        outEquation = MG_Util::ConvertGLEnumToBlendEquation(mode);
+        if (outEquation != ::MobileGL::BlendEquation::Unknown) return true;
+
+        MG_State::pGLContext->RecordError(
+            ErrorCode::InvalidEnum,
+            MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                         "Blend equation enum " + MG_Util::ConvertGLEnumToString(mode) +
+                                             " is not supported."));
+        return false;
+    }
+
+    static Bool TryDecodeStencilFace(GLenum face, const char* functionName, Bool& applyFront, Bool& applyBack) {
+        switch (face) {
+        case GL_FRONT:
+            applyFront = true;
+            applyBack = false;
+            return true;
+        case GL_BACK:
+            applyFront = false;
+            applyBack = true;
+            return true;
+        case GL_FRONT_AND_BACK:
+            applyFront = true;
+            applyBack = true;
+            return true;
+        default:
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                             "Stencil face enum " + MG_Util::ConvertGLEnumToString(face) +
+                                                 " is not supported."));
+            return false;
+        }
+    }
+
+    static Bool TryConvertStencilOperation(GLenum value, const char* functionName, const char* paramName,
+                                           StencilOperation& outOperation) {
+        outOperation = MG_Util::ConvertGLEnumToStencilOperation(value);
+        if (outOperation != StencilOperation::Unknown) return true;
+
+        MG_State::pGLContext->RecordError(
+            ErrorCode::InvalidEnum,
+            MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", functionName,
+                                         String(paramName) + " enum " + MG_Util::ConvertGLEnumToString(value) +
+                                             " is not supported."));
+        return false;
+    }
+
     void Viewport_State(GLint x, GLint y, GLsizei width, GLsizei height) {
         if (width < 0 || height < 0) {
             MG_State::pGLContext->RecordError(ErrorCode::InvalidValue,
@@ -26,27 +103,74 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void StencilOpSeparate_State(GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass) {
-        // TODO: implement
+        Bool applyFront = false;
+        Bool applyBack = false;
+        if (!TryDecodeStencilFace(face, "StencilOpSeparate_State", applyFront, applyBack)) return;
+
+        StencilOperation failOp = StencilOperation::Unknown;
+        StencilOperation depthFailOp = StencilOperation::Unknown;
+        StencilOperation depthPassOp = StencilOperation::Unknown;
+        if (!TryConvertStencilOperation(sfail, "StencilOpSeparate_State", "sfail", failOp) ||
+            !TryConvertStencilOperation(dpfail, "StencilOpSeparate_State", "dpfail", depthFailOp) ||
+            !TryConvertStencilOperation(dppass, "StencilOpSeparate_State", "dppass", depthPassOp)) {
+            return;
+        }
+
+        if (applyFront) {
+            MG_State::pGLContext->SetStencilOp(StencilFace::Front, failOp, depthFailOp, depthPassOp);
+        }
+        if (applyBack) {
+            MG_State::pGLContext->SetStencilOp(StencilFace::Back, failOp, depthFailOp, depthPassOp);
+        }
     }
 
     void StencilOp_State(GLenum fail, GLenum zfail, GLenum zpass) {
-        // TODO: implement
+        StencilOpSeparate_State(GL_FRONT_AND_BACK, fail, zfail, zpass);
     }
 
     void StencilMaskSeparate_State(GLenum face, GLuint mask) {
-        // TODO: implement
+        Bool applyFront = false;
+        Bool applyBack = false;
+        if (!TryDecodeStencilFace(face, "StencilMaskSeparate_State", applyFront, applyBack)) return;
+
+        if (applyFront) {
+            MG_State::pGLContext->SetStencilMask(StencilFace::Front, mask);
+        }
+        if (applyBack) {
+            MG_State::pGLContext->SetStencilMask(StencilFace::Back, mask);
+        }
     }
 
     void StencilMask_State(GLuint mask) {
-        // TODO: implement
+        StencilMaskSeparate_State(GL_FRONT_AND_BACK, mask);
     }
 
     void StencilFuncSeparate_State(GLenum face, GLenum func, GLint ref, GLuint mask) {
-        // TODO: implement
+        Bool applyFront = false;
+        Bool applyBack = false;
+        if (!TryDecodeStencilFace(face, "StencilFuncSeparate_State", applyFront, applyBack)) return;
+
+        DepthTestFunc depthFunc = MG_Util::ConvertGLEnumToDepthTestFunc(func);
+        if (depthFunc == DepthTestFunc::Unknown) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "StencilFuncSeparate_State",
+                                             "Stencil func enum " + MG_Util::ConvertGLEnumToString(func) +
+                                                 " is not supported."));
+            return;
+        }
+
+        const Int clampedRef = std::max(ref, 0);
+        if (applyFront) {
+            MG_State::pGLContext->SetStencilFunc(StencilFace::Front, depthFunc, clampedRef, mask);
+        }
+        if (applyBack) {
+            MG_State::pGLContext->SetStencilFunc(StencilFace::Back, depthFunc, clampedRef, mask);
+        }
     }
 
     void StencilFunc_State(GLenum func, GLint ref, GLuint mask) {
-        // TODO: implement
+        StencilFuncSeparate_State(GL_FRONT_AND_BACK, func, ref, mask);
     }
 
     void Scissor_State(GLint x, GLint y, GLsizei width, GLsizei height) {
@@ -61,27 +185,111 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void SampleCoverage_State(GLfloat value, GLboolean invert) {
-        // TODO: implement
+        MG_State::pGLContext->SetSampleCoverage(std::clamp(static_cast<Float>(value), 0.0f, 1.0f), invert == GL_TRUE);
     }
 
     void PolygonOffset_State(GLfloat factor, GLfloat units) {
-        // TODO: implement
+        MG_State::pGLContext->SetPolygonOffset(static_cast<Float>(factor), static_cast<Float>(units));
     }
 
     void PolygonMode_State(GLenum face, GLenum mode) {
-        // TODO: implement
+        // GL 3.3 core: separate front/back polygon modes were removed in 3.1, so the only legal
+        // face is GL_FRONT_AND_BACK. GL_FRONT / GL_BACK must be rejected (some desktop drivers
+        // leniently accept them, but that is non-conformant). Both errors are GL_INVALID_ENUM and
+        // leave state untouched.
+        if (face != GL_FRONT_AND_BACK) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "glPolygonMode face must be GL_FRONT_AND_BACK in the core profile; got " +
+                                                 MG_Util::ConvertGLEnumToString(face) + "."));
+            return;
+        }
+        if (mode != GL_POINT && mode != GL_LINE && mode != GL_FILL) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "glPolygonMode mode must be GL_POINT, GL_LINE, or GL_FILL; got " +
+                                                 MG_Util::ConvertGLEnumToString(mode) + "."));
+            return;
+        }
+        // Core sets both faces together; keep two slots so GL_POLYGON_MODE round-trips its two values.
+        MG_State::pGLContext->SetPolygonMode(mode, mode);
     }
 
     void PointSize_State(GLfloat size) {
-        // TODO: implement
+        if (size <= 0.0f) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "PointSize_State",
+                                             "Point size must be greater than zero."));
+            return;
+        }
+
+        MG_State::pGLContext->SetPointSize(static_cast<Float>(size));
     }
 
-    void PointParameterf_State(GLenum pname, GLfloat param) {
-        // TODO: implement
+    // Single funnel for all four glPointParameter forms. The two GL 3.3 core pnames both carry one
+    // component, so the *v forms pass params[0], and the integer forms widen to float. The
+    // GL_POINT_SPRITE_COORD_ORIGIN value is an enum passed as a float (36001.0/36002.0 are exact).
+    void PointParameter_State(GLenum pname, GLfloat value) {
+        switch (pname) {
+        case GL_POINT_FADE_THRESHOLD_SIZE:
+            if (value < 0.0f) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                                 "GL_POINT_FADE_THRESHOLD_SIZE must be non-negative."));
+                return;
+            }
+            MG_State::pGLContext->SetPointFadeThresholdSize(value);
+            return;
+        case GL_POINT_SPRITE_COORD_ORIGIN: {
+            const GLenum origin = static_cast<GLenum>(std::lround(value));
+            if (origin != GL_LOWER_LEFT && origin != GL_UPPER_LEFT) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidEnum,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                                 "GL_POINT_SPRITE_COORD_ORIGIN must be GL_LOWER_LEFT or "
+                                                 "GL_UPPER_LEFT."));
+                return;
+            }
+            MG_State::pGLContext->SetPointSpriteCoordOrigin(origin);
+            return;
+        }
+        default:
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Unsupported point parameter pname: " + std::to_string(pname)));
+            return;
+        }
     }
 
+    void PointParameterf_State(GLenum pname, GLfloat param) { PointParameter_State(pname, param); }
+
     void PointParameteri_State(GLenum pname, GLint param) {
-        // TODO: implement
+        PointParameter_State(pname, static_cast<GLfloat>(param));
+    }
+
+    void PointParameterfv_State(GLenum pname, const GLfloat* params) {
+        if (!params) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "params pointer cannot be null."));
+            return;
+        }
+        PointParameter_State(pname, params[0]);
+    }
+
+    void PointParameteriv_State(GLenum pname, const GLint* params) {
+        if (!params) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__, "params pointer cannot be null."));
+            return;
+        }
+        PointParameter_State(pname, static_cast<GLfloat>(params[0]));
     }
 
     void PixelStorei_State(GLenum pname, GLint param) {
@@ -101,14 +309,36 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void LogicOp_State(GLenum opcode) {
-        // TODO: implement
+        LogicOperation logicOp = MG_Util::ConvertGLEnumToLogicOperation(opcode);
+        if (logicOp == LogicOperation::Unknown) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "LogicOp_State",
+                                             "Logic op enum " + MG_Util::ConvertGLEnumToString(opcode) +
+                                                 " is not supported."));
+            return;
+        }
+
+        MG_State::pGLContext->SetLogicOp(logicOp);
     }
 
     void LineWidth_State(GLfloat width) {
-        // TODO: implement
+        if (width <= 0.0f) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "LineWidth_State",
+                                             "Line width must be greater than zero."));
+            return;
+        }
+
+        MG_State::pGLContext->SetLineWidth(static_cast<Float>(width));
     }
 
     GLboolean IsEnabledi_State(GLenum target, GLuint index) {
+        if (!ValidateIndexedBlendCapability(target, index, "IsEnabledi_State")) {
+            return GL_FALSE;
+        }
+
         CapabilityInput capInput = MG_Util::ConvertGLEnumToCapabilityInput(target);
         if (capInput == CapabilityInput::Unknown) {
             MG_State::pGLContext->RecordError(
@@ -122,6 +352,37 @@ namespace MobileGL::MG_Impl::GLImpl {
         return MG_State::pGLContext->IsCapabilityEnabledIndexed(capInput, index) ? GL_TRUE : GL_FALSE;
     }
 
+    void GetBooleani_v_State(GLenum target, GLuint index, GLboolean* data) {
+        if (!data) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "GetBooleani_v_State",
+                                             "data pointer cannot be null."));
+            return;
+        }
+
+        if (target == GL_COLOR_WRITEMASK) {
+            // Indexed per-draw-buffer color writemask writes 4 booleans (R,G,B,A) for draw buffer
+            // `index`. The non-indexed glGetBooleanv(GL_COLOR_WRITEMASK) reports draw buffer 0.
+            if (index >= MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "GetBooleani_v_State",
+                                                 "Color writemask draw buffer index " + std::to_string(index) +
+                                                     " is out of range."));
+                return;
+            }
+            const BoolVec4 mask = MG_State::pGLContext->GetColorMaskIndexed(index);
+            data[0] = mask.x() ? GL_TRUE : GL_FALSE;
+            data[1] = mask.y() ? GL_TRUE : GL_FALSE;
+            data[2] = mask.z() ? GL_TRUE : GL_FALSE;
+            data[3] = mask.w() ? GL_TRUE : GL_FALSE;
+            return;
+        }
+
+        *data = IsEnabledi_State(target, index);
+    }
+
     GLboolean IsEnabled_State(GLenum cap) {
         CapabilityInput capInput = MG_Util::ConvertGLEnumToCapabilityInput(cap);
         if (capInput == CapabilityInput::Unknown) {
@@ -137,14 +398,70 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void Hint_State(GLenum target, GLenum mode) {
-        // TODO: implement
+        switch (target) {
+        case GL_LINE_SMOOTH_HINT:
+        case GL_POLYGON_SMOOTH_HINT:
+        case GL_TEXTURE_COMPRESSION_HINT:
+        case GL_FRAGMENT_SHADER_DERIVATIVE_HINT:
+            break;
+        default:
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Unsupported hint target: " + std::to_string(target)));
+            return;
+        }
+        if (mode != GL_FASTEST && mode != GL_NICEST && mode != GL_DONT_CARE) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "Hint mode must be GL_FASTEST, GL_NICEST or GL_DONT_CARE."));
+            return;
+        }
+        MG_State::pGLContext->SetHint(target, mode);
     }
 
     void FrontFace_State(GLenum mode) {
-        // TODO: implement
+        FrontFaceMode frontFaceMode = MG_Util::ConvertGLEnumToFrontFaceMode(mode);
+        if (frontFaceMode == FrontFaceMode::Unknown) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "FrontFace_State",
+                                             "Front face mode enum " +
+                                                 MG_Util::ConvertFrontFaceModeToString(frontFaceMode) + "(" +
+                                                 MG_Util::ConvertGLEnumToString(mode) + ") is not supported."));
+            return;
+        }
+
+        MG_State::pGLContext->SetFrontFaceMode(frontFaceMode);
+    }
+
+    void ProvokingVertex_State(GLenum mode) {
+        ProvokingVertexMode provokingVertexMode = MG_Util::ConvertGLEnumToProvokingVertexMode(mode);
+        if (provokingVertexMode == ProvokingVertexMode::Unknown) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ProvokingVertex_State",
+                                             "Provoking vertex mode enum " +
+                                                 MG_Util::ConvertProvokingVertexModeToString(provokingVertexMode) +
+                                                 "(" + MG_Util::ConvertGLEnumToString(mode) + ") is not supported."));
+            return;
+        }
+
+        MG_State::pGLContext->SetProvokingVertexMode(provokingVertexMode);
     }
 
     void Enable_State(GLenum cap) {
+        switch (cap) {
+        case GL_TEXTURE_1D:
+        case GL_TEXTURE_2D:
+        case GL_TEXTURE_3D:
+        case GL_TEXTURE_CUBE_MAP:
+            return;
+        default:
+            break;
+        }
+
         CapabilityInput capInput = MG_Util::ConvertGLEnumToCapabilityInput(cap);
         if (capInput == CapabilityInput::Unknown) {
             MG_State::pGLContext->RecordError(
@@ -159,6 +476,16 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void Disable_State(GLenum cap) {
+        switch (cap) {
+        case GL_TEXTURE_1D:
+        case GL_TEXTURE_2D:
+        case GL_TEXTURE_3D:
+        case GL_TEXTURE_CUBE_MAP:
+            return;
+        default:
+            break;
+        }
+
         CapabilityInput capInput = MG_Util::ConvertGLEnumToCapabilityInput(cap);
         if (capInput == CapabilityInput::Unknown) {
             MG_State::pGLContext->RecordError(
@@ -173,7 +500,8 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void DepthRange_State(GLclampd near_val, GLclampd far_val) {
-        // TODO: implement
+        MG_State::pGLContext->SetDepthRange(
+            FloatVec2(ClampUnitFloat(static_cast<GLfloat>(near_val)), ClampUnitFloat(static_cast<GLfloat>(far_val))));
     }
 
     void DepthMask_State(GLboolean flag) {
@@ -210,12 +538,59 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void ColorMask_State(GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha) {
+        // glColorMask broadcasts to every draw buffer. GLboolean coercion: any nonzero value enables
+        // the component; only exactly GL_FALSE disables it.
         MG_State::pGLContext->SetColorMask(
-            BoolVec4(red == GL_TRUE, green == GL_TRUE, blue == GL_TRUE, alpha == GL_TRUE));
+            BoolVec4(red != GL_FALSE, green != GL_FALSE, blue != GL_FALSE, alpha != GL_FALSE));
+    }
+
+    void ColorMaski_State(GLuint buf, GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha) {
+        // Indexed color writemask for the single draw buffer `buf`. buf is a GLuint index, never an
+        // enum, so the only error is GL_INVALID_VALUE when it is out of range (mirrors the indexed
+        // blend entry points, which bound against MAX_DRAW_BUFFERS).
+        if (buf >= MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl", __func__,
+                    "glColorMaski buffer index " + std::to_string(buf) + " is out of range. Max supported is " +
+                        std::to_string(MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS - 1) + "."));
+            return;
+        }
+        MG_State::pGLContext->SetColorMaskIndexed(
+            buf, BoolVec4(red != GL_FALSE, green != GL_FALSE, blue != GL_FALSE, alpha != GL_FALSE));
+    }
+
+    void PrimitiveRestartIndex_State(GLuint index) {
+        // glPrimitiveRestartIndex accepts any GLuint and generates no error.
+        MG_State::pGLContext->SetPrimitiveRestartIndex(index);
     }
 
     void ClampColor_State(GLenum target, GLenum clamp) {
-        // TODO: implement
+        // GL 3.3 core: the only legal target is GL_CLAMP_READ_COLOR. The compatibility-only
+        // GL_CLAMP_VERTEX_COLOR / GL_CLAMP_FRAGMENT_COLOR were removed from the core profile and
+        // must be rejected.
+        if (target != GL_CLAMP_READ_COLOR) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "glClampColor target must be GL_CLAMP_READ_COLOR in the core profile; "
+                                             "got " +
+                                                 MG_Util::ConvertGLEnumToString(target) + "."));
+            return;
+        }
+        // clamp must be one of GL_TRUE, GL_FALSE, or GL_FIXED_ONLY. NOTE: the Khronos man page's
+        // Errors section wrongly omits GL_FIXED_ONLY, but the spec lists it as legal AND it is the
+        // default value, so it must be accepted here.
+        if (clamp != GL_TRUE && clamp != GL_FALSE && clamp != GL_FIXED_ONLY) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", __func__,
+                                             "glClampColor clamp must be GL_TRUE, GL_FALSE, or GL_FIXED_ONLY; got " +
+                                                 MG_Util::ConvertGLEnumToString(clamp) + "."));
+            return;
+        }
+        MG_State::pGLContext->SetClampReadColor(clamp);
     }
 
     void BlendFuncSeparate_State(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha) {
@@ -249,11 +624,24 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void BlendEquation_State(GLenum mode) {
-        // TODO: implement
+        ::MobileGL::BlendEquation blendEquation = ::MobileGL::BlendEquation::Unknown;
+        if (!TryConvertBlendEquation(mode, "BlendEquation_State", blendEquation)) return;
+        MG_State::pGLContext->SetBlendEquation(blendEquation, blendEquation);
+    }
+
+    void BlendEquationSeparate_State(GLenum modeRGB, GLenum modeAlpha) {
+        ::MobileGL::BlendEquation colorEquation = ::MobileGL::BlendEquation::Unknown;
+        if (!TryConvertBlendEquation(modeRGB, "BlendEquationSeparate_State", colorEquation)) return;
+
+        ::MobileGL::BlendEquation alphaEquation = ::MobileGL::BlendEquation::Unknown;
+        if (!TryConvertBlendEquation(modeAlpha, "BlendEquationSeparate_State", alphaEquation)) return;
+
+        MG_State::pGLContext->SetBlendEquation(colorEquation, alphaEquation);
     }
 
     void BlendColor_State(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha) {
-        // TODO: implement
+        MG_State::pGLContext->SetBlendColor(
+            FloatVec4(ClampUnitFloat(red), ClampUnitFloat(green), ClampUnitFloat(blue), ClampUnitFloat(alpha)));
     }
 
     void ClearStencil_State(GLint s) {
@@ -261,7 +649,9 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void ClearDepth_State(GLclampd depth) {
-        MG_State::pGLContext->SetClearDepth(static_cast<Float>(depth));
+        // GL 3.3 §4.2.3: the clear depth is clamped to [0,1] at specification time (Vulkan clear
+        // values additionally require it: VUID-VkClearDepthStencilValue-depth-00022).
+        MG_State::pGLContext->SetClearDepth(ClampUnitFloat(static_cast<Float>(depth)));
     }
 
     void ClearColor_State(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha) {
@@ -283,10 +673,50 @@ namespace MobileGL::MG_Impl::GLImpl {
         BlendFactor dstRGBM = MG_Util::ConvertGLEnumToBlendFactor(dstRGB);
         BlendFactor srcAlphaM = MG_Util::ConvertGLEnumToBlendFactor(srcAlpha);
         BlendFactor dstAlphaM = MG_Util::ConvertGLEnumToBlendFactor(dstAlpha);
+        if (srcRGBM == BlendFactor::Unknown || dstRGBM == BlendFactor::Unknown ||
+            srcAlphaM == BlendFactor::Unknown || dstAlphaM == BlendFactor::Unknown) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidEnum,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "BlendFuncSeparatei_State",
+                                             "One of the indexed blend factor enums is not supported."));
+            return;
+        }
         MG_State::pGLContext->SetBlendFuncIndexed(buf, srcRGBM, dstRGBM, srcAlphaM, dstAlphaM);
     }
 
+    void BlendFunci_State(GLuint buf, GLenum src, GLenum dst) {
+        BlendFuncSeparatei_State(buf, src, dst, src, dst);
+    }
+
+    void BlendEquationSeparatei_State(GLuint buf, GLenum modeRGB, GLenum modeAlpha) {
+        if (buf >= MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue,
+                MakeUnique<GenericErrorInfo>(
+                    "MG_Impl/GLImpl", "BlendEquationSeparatei_State",
+                    "Buffer index " + std::to_string(buf) + " is out of range. Max supported is " +
+                        std::to_string(MG_State::GLState::FramebufferObject::MAX_DRAW_BUFFERS - 1) + "."));
+            return;
+        }
+
+        ::MobileGL::BlendEquation colorEquation = ::MobileGL::BlendEquation::Unknown;
+        if (!TryConvertBlendEquation(modeRGB, "BlendEquationSeparatei_State", colorEquation)) return;
+
+        ::MobileGL::BlendEquation alphaEquation = ::MobileGL::BlendEquation::Unknown;
+        if (!TryConvertBlendEquation(modeAlpha, "BlendEquationSeparatei_State", alphaEquation)) return;
+
+        MG_State::pGLContext->SetBlendEquationIndexed(buf, colorEquation, alphaEquation);
+    }
+
+    void BlendEquationi_State(GLuint buf, GLenum mode) {
+        BlendEquationSeparatei_State(buf, mode, mode);
+    }
+
     void Disablei_State(GLenum target, GLuint index) {
+        if (!ValidateIndexedBlendCapability(target, index, "Disablei_State")) {
+            return;
+        }
+
         auto capInput = MG_Util::ConvertGLEnumToCapabilityInput(target);
         if (capInput == CapabilityInput::Unknown) {
             MG_State::pGLContext->RecordError(
@@ -301,6 +731,10 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void Enablei_State(GLenum target, GLuint index) {
+        if (!ValidateIndexedBlendCapability(target, index, "Enablei_State")) {
+            return;
+        }
+
         auto capInput = MG_Util::ConvertGLEnumToCapabilityInput(target);
         if (capInput == CapabilityInput::Unknown) {
             MG_State::pGLContext->RecordError(
@@ -315,10 +749,26 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     /* @INSERTION_POINT:FUNCTION_IMPLEMENTATION@ */
+    void BlendEquationi(GLuint buf, GLenum mode) {
+        BlendEquationi_State(buf, mode);
+    }
+
+    void BlendEquationSeparatei(GLuint buf, GLenum modeRGB, GLenum modeAlpha) {
+        BlendEquationSeparatei_State(buf, modeRGB, modeAlpha);
+    }
+
+    void BlendFunci(GLuint buf, GLenum src, GLenum dst) {
+        BlendFunci_State(buf, src, dst);
+    }
+
     void BlendFuncSeparatei(GLuint buf, GLenum srcRGB, GLenum dstRGB, GLenum srcAlpha, GLenum dstAlpha) {
         BlendFuncSeparatei_State(buf, srcRGB, dstRGB, srcAlpha, dstAlpha);
     }
 
+    void GetBooleani_v(GLenum target, GLuint index, GLboolean* data) {
+        GetBooleani_v_State(target, index, data);
+    }
+
     void Disablei(GLenum target, GLuint index) {
         Disablei_State(target, index);
     }
@@ -387,10 +837,36 @@ namespace MobileGL::MG_Impl::GLImpl {
         PointParameteri_State(pname, param);
     }
 
+    void PointParameterfv(GLenum pname, const GLfloat* params) {
+        PointParameterfv_State(pname, params);
+    }
+
+    void PointParameteriv(GLenum pname, const GLint* params) {
+        PointParameteriv_State(pname, params);
+    }
+
     void PixelStorei(GLenum pname, GLint param) {
         PixelStorei_State(pname, param);
     }
 
+    void PixelStoref(GLenum pname, GLfloat param) {
+        // Boolean pixel-store pnames convert by a zero-test (0.4 -> TRUE); integer pnames round to
+        // nearest. Branch before converting so a fractional value cannot round a true flag to false.
+        GLint intParam;
+        switch (pname) {
+        case GL_PACK_SWAP_BYTES:
+        case GL_UNPACK_SWAP_BYTES:
+        case GL_PACK_LSB_FIRST:
+        case GL_UNPACK_LSB_FIRST:
+            intParam = (param != 0.0f) ? 1 : 0;
+            break;
+        default:
+            intParam = static_cast<GLint>(std::lround(param));
+            break;
+        }
+        PixelStorei_State(pname, intParam);
+    }
+
     void LogicOp(GLenum opcode) {
         LogicOp_State(opcode);
     }
@@ -415,6 +891,10 @@ namespace MobileGL::MG_Impl::GLImpl {
         FrontFace_State(mode);
     }
 
+    void ProvokingVertex(GLenum mode) {
+        ProvokingVertex_State(mode);
+    }
+
     void Enable(GLenum cap) {
         Enable_State(cap);
     }
@@ -443,10 +923,18 @@ namespace MobileGL::MG_Impl::GLImpl {
         ColorMask_State(red, green, blue, alpha);
     }
 
+    void ColorMaski(GLuint index, GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha) {
+        ColorMaski_State(index, red, green, blue, alpha);
+    }
+
     void ClampColor(GLenum target, GLenum clamp) {
         ClampColor_State(target, clamp);
     }
 
+    void PrimitiveRestartIndex(GLuint index) {
+        PrimitiveRestartIndex_State(index);
+    }
+
     void BlendFuncSeparate(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha) {
         BlendFuncSeparate_State(sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha);
     }
@@ -455,6 +943,10 @@ namespace MobileGL::MG_Impl::GLImpl {
         BlendEquation_State(mode);
     }
 
+    void BlendEquationSeparate(GLenum modeRGB, GLenum modeAlpha) {
+        BlendEquationSeparate_State(modeRGB, modeAlpha);
+    }
+
     void BlendColor(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha) {
         BlendColor_State(red, green, blue, alpha);
     }

MobileGL/MG_Impl/GLImpl/RenderState/GL_RenderState.h

@@ -11,7 +11,11 @@
 
 namespace MobileGL::MG_Impl::GLImpl {
     /* @INSERTION_POINT:FUNCTION_DECLARATION@ */
+    void BlendEquationi(GLuint buf, GLenum mode);
+    void BlendEquationSeparatei(GLuint buf, GLenum modeRGB, GLenum modeAlpha);
+    void BlendFunci(GLuint buf, GLenum src, GLenum dst);
     void BlendFuncSeparatei(GLuint buf, GLenum srcRGB, GLenum dstRGB, GLenum srcAlpha, GLenum dstAlpha);
+    void GetBooleani_v(GLenum target, GLuint index, GLboolean* data);
     void Disablei(GLenum target, GLuint index);
     void Enablei(GLenum target, GLuint index);
     void BlendFunc(GLenum sfactor, GLenum dfactor);
@@ -29,13 +33,17 @@ namespace MobileGL::MG_Impl::GLImpl {
     void PointSize(GLfloat size);
     void PointParameterf(GLenum pname, GLfloat param);
     void PointParameteri(GLenum pname, GLint param);
+    void PointParameterfv(GLenum pname, const GLfloat* params);
+    void PointParameteriv(GLenum pname, const GLint* params);
     void PixelStorei(GLenum pname, GLint param);
+    void PixelStoref(GLenum pname, GLfloat param);
     void LogicOp(GLenum opcode);
     void LineWidth(GLfloat width);
     GLboolean IsEnabledi(GLenum target, GLuint index);
     GLboolean IsEnabled(GLenum cap);
     void Hint(GLenum target, GLenum mode);
     void FrontFace(GLenum mode);
+    void ProvokingVertex(GLenum mode);
     void Enable(GLenum cap);
     void Disable(GLenum cap);
     void DepthRange(GLclampd near_val, GLclampd far_val);
@@ -43,9 +51,12 @@ namespace MobileGL::MG_Impl::GLImpl {
     void DepthFunc(GLenum func);
     void CullFace(GLenum mode);
     void ColorMask(GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha);
+    void ColorMaski(GLuint index, GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha);
     void ClampColor(GLenum target, GLenum clamp);
+    void PrimitiveRestartIndex(GLuint index);
     void BlendFuncSeparate(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
     void BlendEquation(GLenum mode);
+    void BlendEquationSeparate(GLenum modeRGB, GLenum modeAlpha);
     void BlendColor(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
     void ClearStencil(GLint s);
     void ClearDepth(GLclampd depth);

MobileGL/MG_Impl/GLImpl/Sampler/GL_Sampler.cpp

@@ -13,7 +13,45 @@
 #include <MG_Util/Converters/MGToGL/TextureEnumConverter.h>
 
 namespace MobileGL::MG_Impl::GLImpl {
-    void SetSamplerParam_State(GLuint sampler, GLenum pname, const void* param, bool isFloat, bool isInteger) {
+    namespace {
+        Float ReadSamplerScalar(const void* param, Bool isFloat, Bool isUnsignedInteger) {
+            if (isFloat) return *(const GLfloat*)param;
+            if (isUnsignedInteger) return static_cast<Float>(*(const GLuint*)param);
+            return static_cast<Float>(*(const GLint*)param);
+        }
+
+        Bool ValidateSamplerParameterValue(GLenum pname, const void* param, Bool isFloat, Bool isUnsignedInteger) {
+            if (param == nullptr) return false;
+
+            switch (pname) {
+            case GL_TEXTURE_MIN_LOD:
+            case GL_TEXTURE_MAX_LOD:
+            case GL_TEXTURE_LOD_BIAS:
+                return true;
+            case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+                if (ReadSamplerScalar(param, isFloat, isUnsignedInteger) >= 1.0f) return true;
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "SetSamplerParam_State",
+                                                 "GL_TEXTURE_MAX_ANISOTROPY_EXT must be at least 1.0."));
+                return false;
+            default:
+                break;
+            }
+
+            if (isFloat) {
+                return SamplerImpl::ValidateSamplerFloatParam(pname, *(const GLfloat*)param);
+            }
+            if (isUnsignedInteger) {
+                return SamplerImpl::ValidateSamplerIntParam(pname, static_cast<GLint>(*(const GLuint*)param));
+            }
+            return SamplerImpl::ValidateSamplerIntParam(pname, *(const GLint*)param);
+        }
+    } // namespace
+
+    void SetSamplerParam_State(GLuint sampler, GLenum pname, const void* param, bool isFloat,
+                               bool isUnsignedInteger) {
+        if (param == nullptr) return;
         if (!SamplerImpl::ValidateSamplerName(sampler)) return;
 
         Bool doesSamplerObjectCreated = MG_State::pGLContext->ValidateSamplerObject(sampler);
@@ -23,6 +61,7 @@ namespace MobileGL::MG_Impl::GLImpl {
         }
         auto& samplerObj = MG_State::pGLContext->GetSamplerObject(sampler);
         if (!SamplerImpl::ValidateSamplerObject(sampler)) return;
+        if (!ValidateSamplerParameterValue(pname, param, isFloat, isUnsignedInteger)) return;
 
         using namespace MG_Util;
         switch (pname) {
@@ -51,6 +90,9 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_TEXTURE_LOD_BIAS:
             samplerObj->SetLodBias(*(const GLfloat*)param);
             break;
+        case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+            samplerObj->SetMaxAnisotropy(ReadSamplerScalar(param, isFloat, isUnsignedInteger));
+            break;
         case GL_TEXTURE_COMPARE_MODE:
             samplerObj->SetCompareMode(MG_Util::ConvertGLEnumToSamplerCompareMode(*(const GLint*)param));
             break;
@@ -64,7 +106,9 @@ namespace MobileGL::MG_Impl::GLImpl {
         }
     }
 
-    void GetSamplerParam_State(GLuint sampler, GLenum pname, void* params, bool isFloat, bool isInteger) {
+    void GetSamplerParam_State(GLuint sampler, GLenum pname, void* params, bool isFloat,
+                               bool isUnsignedInteger) {
+        if (params == nullptr) return;
         if (!SamplerImpl::ValidateSamplerName(sampler)) return;
 
         Bool doesSamplerObjectCreated = MG_State::pGLContext->ValidateSamplerObject(sampler);
@@ -103,6 +147,15 @@ namespace MobileGL::MG_Impl::GLImpl {
         case GL_TEXTURE_LOD_BIAS:
             *(GLfloat*)params = samplerObj->GetLodBias();
             break;
+        case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+            if (isFloat) {
+                *(GLfloat*)params = samplerObj->GetMaxAnisotropy();
+            } else if (isUnsignedInteger) {
+                *(GLuint*)params = static_cast<GLuint>(samplerObj->GetMaxAnisotropy());
+            } else {
+                *(GLint*)params = static_cast<GLint>(samplerObj->GetMaxAnisotropy());
+            }
+            break;
         case GL_TEXTURE_COMPARE_MODE:
             *(GLuint*)params = MG_Util::ConvertSamplerCompareModeToGLEnum(samplerObj->GetCompareMode());
             break;
@@ -167,6 +220,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void BindSampler_State(GLuint unit, GLuint sampler) {
+        MGLOG_D("BindSampler_State: unit = %u, sampler = %u", unit, sampler);
         if (unit >= MG_State::GLState::TextureState::MAX_TEXTURE_IMAGE_UNITS) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue,
@@ -175,10 +229,20 @@ namespace MobileGL::MG_Impl::GLImpl {
         }
 
         auto& textureUnit = MG_State::pGLContext->GetTextureUnitObject((Int)unit);
+        MG_State::pGLContext->NoteTextureUnitTouched((Int)unit);
         if (sampler == 0) {
             textureUnit.SetSamplerObject(nullptr);
         } else {
-            if (!SamplerImpl::ValidateSamplerName(sampler)) return;
+            // GL 3.3 core 3.8.2: BindSampler on a name GenSamplers never returned - or one already
+            // deleted - is INVALID_OPERATION. SamplerParameter* raises INVALID_VALUE for the same
+            // name, which is why this cannot go through the shared SamplerImpl validator.
+            if (!MG_State::pGLContext->ValidateSamplerName(sampler)) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidOperation,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "BindSampler_State",
+                                                 std::format("Invalid sampler name {}", sampler)));
+                return;
+            }
             Bool doesSamplerObjectCreated = MG_State::pGLContext->ValidateSamplerObject(sampler);
             if (!doesSamplerObjectCreated) {
                 MG_State::pGLContext->CreateSamplerObject(sampler);
@@ -212,7 +276,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void SamplerParameterIiv(GLuint sampler, GLenum pname, const GLint* param) {
-        SetSamplerParam_State(sampler, pname, param, false, true);
+        SetSamplerParam_State(sampler, pname, param, false, false);
     }
 
     void SamplerParameteriv(GLuint sampler, GLenum pname, const GLint* param) {
@@ -240,7 +304,7 @@ namespace MobileGL::MG_Impl::GLImpl {
     }
 
     void GetSamplerParameterIiv(GLuint sampler, GLenum pname, GLint* params) {
-        GetSamplerParam_State(sampler, pname, params, false, true);
+        GetSamplerParam_State(sampler, pname, params, false, false);
     }
 
     void GetSamplerParameterfv(GLuint sampler, GLenum pname, GLfloat* params) {

MobileGL/MG_Impl/GLImpl/Sampler/Validators.cpp

@@ -99,6 +99,16 @@ namespace MobileGL::MG_Impl::GLImpl::SamplerImpl {
         case GL_TEXTURE_LOD_BIAS:
             return true;
 
+        case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+            if (!(param >= 1.0f)) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateSamplerFloatParam",
+                                                 "GL_TEXTURE_MAX_ANISOTROPY_EXT must be at least 1.0."));
+                return false;
+            }
+            return true;
+
         case GL_TEXTURE_BORDER_COLOR:
             if (param < 0.0f || param > 1.0f) {
                 MG_State::pGLContext->RecordError(
@@ -125,6 +135,16 @@ namespace MobileGL::MG_Impl::GLImpl::SamplerImpl {
             }
             return true;
 
+        case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+            if (param < 1) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateSamplerIntParam",
+                                                 "GL_TEXTURE_MAX_ANISOTROPY_EXT must be at least 1."));
+                return false;
+            }
+            return true;
+
         default:
             return ValidateSamplerParam(pname, static_cast<GLenum>(param));
         }

MobileGL/MG_Impl/GLImpl/Sync/GL_Sync.cpp

@@ -7,37 +7,130 @@
 // End of Source File Header
 
 #include "GL_Sync.h"
-
-#include "MG_State/GLState/Core.h"
+#include <MG_Backend/BackendObjects.h>
 
 namespace MobileGL::MG_Impl::GLImpl {
-    GLsync FenceSync_Backend(GLenum condition, GLbitfield flags) {
-        return 0;
-    }
+    namespace {
+        // Frontend sync object: wraps an optional backend fence handle. A null
+        // backend handle (backend has no fence support, or could not create a
+        // fence at call time) keeps the legacy always-signaled behavior.
+        struct SyncObject {
+            MG_Backend::BackendSyncHandle backendHandle = nullptr;
+            GLenum condition = GL_SYNC_GPU_COMMANDS_COMPLETE;
+            GLbitfield flags = 0;
+        };
 
-    GLenum ClientWaitSync_Backend(GLsync sync, GLbitfield flags, GLuint64 timeout) {
-        return 0;
-    }
+        // Sync calls may arrive from any thread (launchers migrate the context
+        // across JVM threads), so the live-object registry is mutex-guarded.
+        // Entries left at process shutdown are simply dropped; their backend
+        // handles die with the backend.
+        std::mutex g_syncObjectsMutex;
+        UnorderedMap<GLsync, SyncObject*> g_liveSyncObjects;
 
-    void DeleteSync_Backend(GLsync sync) {}
-
-    GLsync FenceSync_State(GLenum condition, GLbitfield flags) {
-        return 0;
-    }
-
-    GLenum ClientWaitSync_State(GLsync sync, GLbitfield flags, GLuint64 timeout) {
-        return 0;
-    }
-
-    void DeleteSync_State(GLsync sync) {}
+        SyncObject* FindSyncObject(GLsync sync) {
+            const std::lock_guard<std::mutex> lock(g_syncObjectsMutex);
+            const auto it = g_liveSyncObjects.find(sync);
+            return it != g_liveSyncObjects.end() ? it->second : nullptr;
+        }
+    } // namespace
 
     GLsync FenceSync(GLenum condition, GLbitfield flags) {
-        return 0;
+        auto* syncObject = new SyncObject;
+        syncObject->condition = condition;
+        syncObject->flags = flags;
+        if (const auto backendFenceSync = MG_Backend::gBackendFunctionsTable.GL.FenceSync) {
+            syncObject->backendHandle = backendFenceSync();
+        }
+        const GLsync handle = reinterpret_cast<GLsync>(syncObject);
+        const std::lock_guard<std::mutex> lock(g_syncObjectsMutex);
+        g_liveSyncObjects[handle] = syncObject;
+        return handle;
+    }
+
+    GLboolean IsSync(GLsync sync) {
+        return FindSyncObject(sync) != nullptr ? GL_TRUE : GL_FALSE;
     }
 
     GLenum ClientWaitSync(GLsync sync, GLbitfield flags, GLuint64 timeout) {
-        return 0;
+        const auto* syncObject = FindSyncObject(sync);
+        if (!syncObject) {
+            return GL_WAIT_FAILED;
+        }
+        const auto backendClientWaitSync = MG_Backend::gBackendFunctionsTable.GL.ClientWaitSync;
+        if (!backendClientWaitSync || !syncObject->backendHandle) {
+            return GL_ALREADY_SIGNALED; // legacy always-signaled fallback
+        }
+        return backendClientWaitSync(syncObject->backendHandle, flags, timeout);
     }
 
-    void DeleteSync(GLsync sync) {}
+    void WaitSync(GLsync sync, GLbitfield flags, GLuint64 timeout) {
+        const auto* syncObject = FindSyncObject(sync);
+        if (!syncObject) {
+            return;
+        }
+        const auto backendWaitSync = MG_Backend::gBackendFunctionsTable.GL.WaitSync;
+        if (backendWaitSync && syncObject->backendHandle) {
+            backendWaitSync(syncObject->backendHandle, flags, timeout);
+        }
+    }
+
+    void DeleteSync(GLsync sync) {
+        if (sync == nullptr) {
+            return; // glDeleteSync(0) is silently ignored
+        }
+        SyncObject* syncObject = nullptr;
+        {
+            const std::lock_guard<std::mutex> lock(g_syncObjectsMutex);
+            const auto it = g_liveSyncObjects.find(sync);
+            if (it == g_liveSyncObjects.end()) {
+                return;
+            }
+            syncObject = it->second;
+            g_liveSyncObjects.erase(it);
+        }
+        const auto backendDeleteSync = MG_Backend::gBackendFunctionsTable.GL.DeleteSync;
+        if (backendDeleteSync && syncObject->backendHandle) {
+            backendDeleteSync(syncObject->backendHandle);
+        }
+        delete syncObject;
+    }
+
+    void GetSynciv(GLsync sync, GLenum pname, GLsizei bufSize, GLsizei* length, GLint* values) {
+        const auto* syncObject = FindSyncObject(sync);
+        if (!syncObject) {
+            if (length) {
+                *length = 0;
+            }
+            return;
+        }
+
+        GLint value = 0;
+        switch (pname) {
+        case GL_OBJECT_TYPE:
+            value = GL_SYNC_FENCE;
+            break;
+        case GL_SYNC_STATUS: {
+            const auto backendGetSyncStatus = MG_Backend::gBackendFunctionsTable.GL.GetSyncStatus;
+            const Bool signaled = !backendGetSyncStatus || !syncObject->backendHandle ||
+                                  backendGetSyncStatus(syncObject->backendHandle);
+            value = signaled ? GL_SIGNALED : GL_UNSIGNALED;
+            break;
+        }
+        case GL_SYNC_CONDITION:
+            value = static_cast<GLint>(syncObject->condition);
+            break;
+        case GL_SYNC_FLAGS:
+            value = static_cast<GLint>(syncObject->flags);
+            break;
+        default:
+            break;
+        }
+
+        if (length) {
+            *length = bufSize > 0 && values ? 1 : 0;
+        }
+        if (bufSize > 0 && values) {
+            values[0] = value;
+        }
+    }
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Sync/GL_Sync.h

@@ -11,6 +11,9 @@
 
 namespace MobileGL::MG_Impl::GLImpl {
     GLsync FenceSync(GLenum condition, GLbitfield flags);
+    GLboolean IsSync(GLsync sync);
     GLenum ClientWaitSync(GLsync sync, GLbitfield flags, GLuint64 timeout);
+    void WaitSync(GLsync sync, GLbitfield flags, GLuint64 timeout);
     void DeleteSync(GLsync sync);
+    void GetSynciv(GLsync sync, GLenum pname, GLsizei bufSize, GLsizei* length, GLint* values);
 } // namespace MobileGL::MG_Impl::GLImpl

MobileGL/MG_Impl/GLImpl/Texture/GL_Texture.h

@@ -11,8 +11,53 @@
 
 namespace MobileGL::MG_Impl::GLImpl {
     /* @INSERTION_POINT:FUNCTION_DECLARATION@ */
+    void ClearTexImage(GLuint texture, GLint level, GLenum format, GLenum type, const void* data);
+    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);
+    void BindImageTexture(GLuint unit, GLuint texture, GLint level, GLboolean layered, GLint layer, GLenum access,
+                          GLenum format);
     void GenerateMipmap(GLenum target);
     void GetTexImage(GLenum target, GLint level, GLenum format, GLenum type, GLvoid* pixels);
+    void CreateTextures(GLenum target, GLsizei n, GLuint* textures);
+    void TextureStorage1D(GLuint texture, GLsizei levels, GLenum internalformat, GLsizei width);
+    void TextureStorage2D(GLuint texture, GLsizei levels, GLenum internalformat, GLsizei width, GLsizei height);
+    void TextureStorage3D(GLuint texture, GLsizei levels, GLenum internalformat, GLsizei width, GLsizei height,
+                          GLsizei depth);
+    void TextureStorage2DMultisample(GLuint texture, GLsizei samples, GLenum internalformat, GLsizei width,
+                                     GLsizei height, GLboolean fixedsamplelocations);
+    void TextureStorage3DMultisample(GLuint texture, GLsizei samples, GLenum internalformat, GLsizei width,
+                                     GLsizei height, GLsizei depth, GLboolean fixedsamplelocations);
+    void TextureSubImage1D(GLuint texture, GLint level, GLint xoffset, GLsizei width, GLenum format, GLenum type,
+                           const void* pixels);
+    void TextureSubImage2D(GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height,
+                           GLenum format, GLenum type, const void* pixels);
+    void TextureSubImage3D(GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width,
+                           GLsizei height, GLsizei depth, GLenum format, GLenum type, const void* pixels);
+    void TextureParameterf(GLuint texture, GLenum pname, GLfloat param);
+    void TextureParameterfv(GLuint texture, GLenum pname, const GLfloat* params);
+    void TextureParameteri(GLuint texture, GLenum pname, GLint param);
+    void TextureParameterIiv(GLuint texture, GLenum pname, const GLint* params);
+    void TextureParameterIuiv(GLuint texture, GLenum pname, const GLuint* params);
+    void TextureParameteriv(GLuint texture, GLenum pname, const GLint* params);
+    void GenerateTextureMipmap(GLuint texture);
+    void BindTextureUnit(GLuint unit, GLuint texture);
+    void GetTextureImage(GLuint texture, GLint level, GLenum format, GLenum type, GLsizei bufSize, void* pixels);
+    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);
+    void GetTextureParameterfv(GLuint texture, GLenum pname, GLfloat* params);
+    void GetTextureParameterIiv(GLuint texture, GLenum pname, GLint* params);
+    void GetTextureParameterIuiv(GLuint texture, GLenum pname, GLuint* params);
+    void GetTextureParameteriv(GLuint texture, GLenum pname, GLint* params);
+    void GetTextureLevelParameterfv(GLuint texture, GLint level, GLenum pname, GLfloat* params);
+    void GetTextureLevelParameteriv(GLuint texture, GLint level, GLenum pname, GLint* params);
+    void TexStorage1D(GLenum target, GLsizei levels, GLenum internalformat, GLsizei width);
+    void TexStorage2D(GLenum target, GLsizei levels, GLenum internalformat, GLsizei width, GLsizei height);
+    void TexStorage3D(GLenum target, GLsizei levels, GLenum internalformat, GLsizei width, GLsizei height,
+                      GLsizei depth);
+    void TexStorage2DMultisample(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width,
+                                 GLsizei height, GLboolean fixedsamplelocations);
+    void TexStorage3DMultisample(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width,
+                                 GLsizei height, GLsizei depth, GLboolean fixedsamplelocations);
     void TexSubImage3D(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width,
                        GLsizei height, GLsizei depth, GLenum format, GLenum type, const void* pixels);
     void TexSubImage2D(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height,
@@ -44,6 +89,8 @@ namespace MobileGL::MG_Impl::GLImpl {
     void GetTexParameterfv(GLenum target, GLenum pname, GLfloat* params);
     void GetTexLevelParameteriv(GLenum target, GLint level, GLenum pname, GLint* params);
     void GetTexLevelParameterfv(GLenum target, GLint level, GLenum pname, GLfloat* params);
+    void GetMultisamplefv(GLenum pname, GLuint index, GLfloat* val);
+    void GetInternalformativ(GLenum target, GLenum internalformat, GLenum pname, GLsizei bufSize, GLint* params);
     void GetCompressedTexImage(GLenum target, GLint level, void* img);
     void GenTextures(GLsizei n, GLuint* textures);
     void DeleteTextures(GLsizei n, const GLuint* textures);
@@ -51,11 +98,16 @@ namespace MobileGL::MG_Impl::GLImpl {
                            GLsizei width, GLsizei height);
     void CopyTexSubImage2D(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width,
                            GLsizei height);
+    void CopyTextureSubImage2D(GLuint texture, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y,
+                               GLsizei width, GLsizei height);
     void CopyTexSubImage1D(GLenum target, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width);
     void CopyTexImage2D(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width,
                         GLsizei height, GLint border);
     void CopyTexImage1D(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width,
                         GLint border);
+    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);
     void CompressedTexSubImage3D(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width,
                                  GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const void* data);
     void CompressedTexSubImage2D(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width,

MobileGL/MG_Impl/GLImpl/Texture/ProxyTexture.cpp

@@ -7,10 +7,15 @@
 // End of Source File Header
 
 #include "ProxyTexture.h"
+#include <MG_State/GLState/TextureState/TextureObject1D.h>
 #include <MG_State/GLState/TextureState/TextureObject2D.h>
+#include <MG_State/GLState/TextureState/TextureObject2DCube.h>
+#include <MG_State/GLState/TextureState/TextureObject3D.h>
+#include <MG_State/GLState/TextureState/TextureObjectStubs.h>
 
 namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
-    UniquePtr<ProxyTextureManager> pProxyTextureManager;
+    // Leak-at-exit storage; see GlobalObjects.cpp.
+    UniquePtr<ProxyTextureManager>& pProxyTextureManager = *new UniquePtr<ProxyTextureManager>();
 
     Bool IsProxyTextureTarget(TextureUploadTarget target) {
         switch (target) {
@@ -37,7 +42,39 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
             m_proxyTexturesMap.erase(it);
         }
         auto& obj = m_proxyTexturesMap[target];
-        obj = MakeShared<MG_State::GLState::TextureObject2D>(0);
+        switch (target) {
+        case TextureUploadTarget::ProxyTexture1D:
+        case TextureUploadTarget::ProxyTexture1DArray:
+            obj = MakeShared<MG_State::GLState::TextureObject1D>(0);
+            break;
+        case TextureUploadTarget::ProxyTexture2D:
+            obj = MakeShared<MG_State::GLState::TextureObject2D>(0);
+            break;
+        case TextureUploadTarget::ProxyTexture3D:
+            obj = MakeShared<MG_State::GLState::TextureObject3D>(0);
+            break;
+        case TextureUploadTarget::ProxyCubeMap:
+            obj = MakeShared<MG_State::GLState::TextureObject2DCube>(0);
+            break;
+        case TextureUploadTarget::ProxyTextureRectangle:
+            obj = MakeShared<MG_State::GLState::TextureObjectRectangle>(0);
+            break;
+        case TextureUploadTarget::ProxyTexture2DArray:
+            obj = MakeShared<MG_State::GLState::TextureObject2DArray>(0);
+            break;
+        case TextureUploadTarget::ProxyCubeMapArray:
+            obj = MakeShared<MG_State::GLState::TextureObjectCubeMapArray>(0);
+            break;
+        case TextureUploadTarget::ProxyTexture2DMultisample:
+            obj = MakeShared<MG_State::GLState::TextureObject2DMultisample>(0);
+            break;
+        case TextureUploadTarget::ProxyTexture2DMultisampleArray:
+            obj = MakeShared<MG_State::GLState::TextureObject2DMultisampleArray>(0);
+            break;
+        default:
+            obj = MakeShared<MG_State::GLState::TextureObject2D>(0);
+            break;
+        }
         return obj;
     }
 

MobileGL/MG_Impl/GLImpl/Texture/ProxyTexture.h

@@ -23,5 +23,5 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
         UnorderedMap<TextureUploadTarget, SharedPtr<MG_State::GLState::ITextureObject>> m_proxyTexturesMap;
     };
 
-    extern UniquePtr<ProxyTextureManager> pProxyTextureManager;
+    extern UniquePtr<ProxyTextureManager>& pProxyTextureManager;
 } // namespace MobileGL::MG_Impl::GLImpl::TextureImpl

MobileGL/MG_Impl/GLImpl/Texture/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>
@@ -83,8 +84,20 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
             return false;
         }
 
-        // TODO: GL_INVALID_VALUE may be generated if level is greater than log2(max), where max is the returned
-        // value of GL_MAX_TEXTURE_SIZE.
+        Int maxTextureSize = MG_Backend::DynamicBackendParameters{}.MaxTextureSize;
+        if (MG_Backend::pActiveBackendObject) {
+            maxTextureSize = MG_Backend::pActiveBackendObject->GetDynamicParameters().MaxTextureSize;
+        }
+        Int maxLevel = 0;
+        for (Int size = std::max(maxTextureSize, 1); size > 1; size >>= 1) {
+            ++maxLevel;
+        }
+        if (level > maxLevel) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidValue, MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateTextureLevelNumber",
+                                                                      "Texture level exceeds GL_MAX_TEXTURE_SIZE"));
+            return false;
+        }
 
         return true;
     }
@@ -129,7 +142,7 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
         return true;
     }
 
-    Bool ValidateTextureSizeRange(SizeT width, SizeT height, SizeT depth) {
+    Bool ValidateTextureSizeRange(Int width, Int height, Int depth) {
         if (width < 0 || height < 0 || depth < 0) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidValue, MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateTextureSizeRange",
@@ -162,61 +175,153 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
         return true;
     }
 
-    Bool ValidateTextureInternalFormatCompatibleWithInput(TextureInputFormat format,
-                                                          TextureInternalFormat internalFormat,
-                                                          TexturePixelDataType type) {
-        if (type == TexturePixelDataType::UnsignedByte332 || type == TexturePixelDataType::UnsignedByte233Rev ||
-            type == TexturePixelDataType::UnsignedShort565 || type == TexturePixelDataType::UnsignedShort565Rev ||
-            type == TexturePixelDataType::UnsignedInt101111Rev) {
-            if (format != TextureInputFormat::RGB) {
-                MG_State::pGLContext->RecordError(
-                    ErrorCode::InvalidOperation,
-                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateTextureInternalFormatCompatibleWithInput",
-                                                 "Invalid format for the given type"));
+    Bool IsIntegerColorInputFormat(TextureInputFormat format) {
+        return format == TextureInputFormat::RInteger || format == TextureInputFormat::RGInteger ||
+               format == TextureInputFormat::RGBInteger || format == TextureInputFormat::BGRInteger ||
+               format == TextureInputFormat::RGBAInteger || format == TextureInputFormat::BGRAInteger ||
+               format == TextureInputFormat::GreenInteger || format == TextureInputFormat::BlueInteger ||
+               format == TextureInputFormat::AlphaInteger;
+    }
+
+    Bool IsIntegerColorInternalFormat(TextureInternalFormat internalFormat) {
+        switch (internalFormat) {
+            case TextureInternalFormat::R8I:
+            case TextureInternalFormat::R8UI:
+            case TextureInternalFormat::R16I:
+            case TextureInternalFormat::R16UI:
+            case TextureInternalFormat::R32I:
+            case TextureInternalFormat::R32UI:
+            case TextureInternalFormat::RG8I:
+            case TextureInternalFormat::RG8UI:
+            case TextureInternalFormat::RG16I:
+            case TextureInternalFormat::RG16UI:
+            case TextureInternalFormat::RG32I:
+            case TextureInternalFormat::RG32UI:
+            case TextureInternalFormat::RGB8I:
+            case TextureInternalFormat::RGB8UI:
+            case TextureInternalFormat::RGB16I:
+            case TextureInternalFormat::RGB16UI:
+            case TextureInternalFormat::RGB32I:
+            case TextureInternalFormat::RGB32UI:
+            case TextureInternalFormat::RGBA8I:
+            case TextureInternalFormat::RGBA8UI:
+            case TextureInternalFormat::RGBA16I:
+            case TextureInternalFormat::RGBA16UI:
+            case TextureInternalFormat::RGBA32I:
+            case TextureInternalFormat::RGBA32UI:
+            case TextureInternalFormat::RGB10A2UI:
+                return true;
+            default:
                 return false;
+        }
+    }
+
+    static Bool IsDepthLikeInternalFormat(TextureInternalFormat internalFormat) {
+        switch (internalFormat) {
+            case TextureInternalFormat::DepthComponent:
+            case TextureInternalFormat::DepthComponent16:
+            case TextureInternalFormat::DepthComponent24:
+            case TextureInternalFormat::DepthComponent32: // not core, kept for Minecraft 1.21.5+
+            case TextureInternalFormat::DepthComponent32F:
+            case TextureInternalFormat::Depth24Stencil8:
+            case TextureInternalFormat::Depth32FStencil8:
+            case TextureInternalFormat::DepthStencil:
+                return true;
+            default:
+                return false;
+        }
+    }
+
+    static Bool IsDepthLikeInputFormat(TextureInputFormat format) {
+        return format == TextureInputFormat::DepthComponent || format == TextureInputFormat::DepthStencil ||
+               format == TextureInputFormat::StencilIndex;
+    }
+
+    // Client-memory format<->type pairing rules shared by pixel uploads (TexImage*) and readbacks
+    // (ReadPixels, GetTexImage). Mirrors the desktop-GL validity matrix used by GL CTS packed_pixels
+    // (glcPackedPixelsTests isFormatValid): packed types constrain the formats they may pair with, and
+    // integer formats reject floating-point types; violations raise GL_INVALID_OPERATION.
+    Bool ValidateClientFormatTypePairing(TextureInputFormat format, TexturePixelDataType type) {
+        const auto recordInvalidOperation = [](const char* message) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateClientFormatTypePairing", message));
+            return false;
+        };
+
+        if (type == TexturePixelDataType::UnsignedByte332 || type == TexturePixelDataType::UnsignedByte233Rev ||
+            type == TexturePixelDataType::UnsignedShort565 || type == TexturePixelDataType::UnsignedShort565Rev) {
+            if (format != TextureInputFormat::RGB && format != TextureInputFormat::RGBInteger) {
+                return recordInvalidOperation("Packed RGB type requires RGB or RGB_INTEGER format");
+            }
+        }
+
+        if (type == TexturePixelDataType::UnsignedInt101111Rev || type == TexturePixelDataType::UnsignedInt5999Rev) {
+            if (format != TextureInputFormat::RGB) {
+                return recordInvalidOperation("Packed float RGB type requires RGB format");
             }
         }
 
         if (type == TexturePixelDataType::UnsignedShort4444 || type == TexturePixelDataType::UnsignedShort4444Rev ||
             type == TexturePixelDataType::UnsignedShort5551 || type == TexturePixelDataType::UnsignedShort1555Rev ||
             type == TexturePixelDataType::UnsignedInt8888 || type == TexturePixelDataType::UnsignedInt8888Rev ||
-            type == TexturePixelDataType::UnsignedInt1010102 || type == TexturePixelDataType::UnsignedInt2101010Rev ||
-            type == TexturePixelDataType::UnsignedInt5999Rev) {
-            if (format != TextureInputFormat::RGBA && format != TextureInputFormat::BGRA) {
-                MG_State::pGLContext->RecordError(
-                    ErrorCode::InvalidOperation,
-                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateTextureInternalFormatCompatibleWithInput",
-                                                 "Invalid format for the given type"));
-                return false;
+            type == TexturePixelDataType::UnsignedInt1010102 || type == TexturePixelDataType::UnsignedInt2101010Rev) {
+            if (format != TextureInputFormat::RGBA && format != TextureInputFormat::BGRA &&
+                format != TextureInputFormat::RGBAInteger && format != TextureInputFormat::BGRAInteger) {
+                return recordInvalidOperation("Packed RGBA type requires RGBA/BGRA (integer) format");
             }
         }
 
-        if (internalFormat == TextureInternalFormat::DepthComponent ||
-            internalFormat == TextureInternalFormat::DepthComponent16 ||
-            internalFormat == TextureInternalFormat::DepthComponent24 ||
-            internalFormat == TextureInternalFormat::DepthComponent32F) {
-            if (format != TextureInputFormat::DepthComponent) {
-                MG_State::pGLContext->RecordError(
-                    ErrorCode::InvalidOperation,
-                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateTextureInternalFormatCompatibleWithInput",
-                                                 "Invalid format for depth component internal format"));
-                return false;
+        if (type == TexturePixelDataType::UnsignedInt248 || type == TexturePixelDataType::Float32UnsignedInt248Rev) {
+            if (format != TextureInputFormat::DepthStencil) {
+                return recordInvalidOperation("Packed depth-stencil type requires DEPTH_STENCIL format");
             }
         }
 
-        if (format == TextureInputFormat::DepthComponent &&
-            (internalFormat != TextureInternalFormat::DepthComponent &&
-             internalFormat != TextureInternalFormat::DepthComponent16 &&
-             internalFormat != TextureInternalFormat::DepthComponent24 &&
-             internalFormat != TextureInternalFormat::DepthComponent32F &&
-             internalFormat != TextureInternalFormat::DepthComponent32 // workaround for Minecraft 1.21.5+
-             )) {
+        if (format == TextureInputFormat::DepthStencil && type != TexturePixelDataType::UnsignedInt248 &&
+            type != TexturePixelDataType::Float32UnsignedInt248Rev) {
+            return recordInvalidOperation("DEPTH_STENCIL format requires a packed depth-stencil type");
+        }
+
+        if (IsIntegerColorInputFormat(format) &&
+            (type == TexturePixelDataType::Float || type == TexturePixelDataType::HalfFloat)) {
+            return recordInvalidOperation("Integer format cannot be used with a floating-point type");
+        }
+
+        return true;
+    }
+
+    // Mirrors the desktop-GL validity matrix used by GL CTS packed_pixels (glcPackedPixelsTests
+    // isFormatValid, INPUT_TEXIMAGE): packed-type/format pairing, depth-vs-color mismatch, and
+    // integer-ness matching all raise GL_INVALID_OPERATION instead of reaching the upload path.
+    Bool ValidateTextureInternalFormatCompatibleWithInput(TextureInputFormat format,
+                                                          TextureInternalFormat internalFormat,
+                                                          TexturePixelDataType type) {
+        const auto recordInvalidOperation = [](const char* message) {
             MG_State::pGLContext->RecordError(
                 ErrorCode::InvalidOperation,
                 MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateTextureInternalFormatCompatibleWithInput",
-                                             "Invalid internal format for depth component format"));
+                                             message));
+            return false;
+        };
+
+        if (!ValidateClientFormatTypePairing(format, type)) {
             return false;
         }
+
+        // TexImage in core 3.3 has no stencil-only upload path (that arrived with GL 4.4).
+        if (format == TextureInputFormat::StencilIndex) {
+            return recordInvalidOperation("STENCIL_INDEX is not a valid texture upload format");
+        }
+
+        if (IsDepthLikeInputFormat(format) != IsDepthLikeInternalFormat(internalFormat)) {
+            return recordInvalidOperation("Depth/stencil-ness of format and internal format must match");
+        }
+
+        if (IsIntegerColorInputFormat(format) != IsIntegerColorInternalFormat(internalFormat)) {
+            return recordInvalidOperation("Integer-ness of format and internal format must match");
+        }
+
         return true;
     }
 
@@ -230,6 +335,18 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
                 return false;
             }
         }
+        if (target == TextureUploadTarget::Texture2DMultisample ||
+            target == TextureUploadTarget::ProxyTexture2DMultisample ||
+            target == TextureUploadTarget::Texture2DMultisampleArray ||
+            target == TextureUploadTarget::ProxyTexture2DMultisampleArray) {
+            if (level != 0) {
+                MG_State::pGLContext->RecordError(
+                    ErrorCode::InvalidValue,
+                    MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", "ValidateTextureLevelWithUploadTarget",
+                                                 "Level must be zero for multisample textures"));
+                return false;
+            }
+        }
         return true;
     }
 
@@ -243,6 +360,19 @@ namespace MobileGL::MG_Impl::GLImpl::TextureImpl {
         return true;
     }
 
+    Bool ValidateTextureNotDefault(const SharedPtr<MG_State::GLState::ITextureObject>& textureObject,
+                                   const char* caller) {
+        if (textureObject && textureObject->GetExternalIndex() == 0) {
+            MG_State::pGLContext->RecordError(
+                ErrorCode::InvalidOperation,
+                MakeUnique<GenericErrorInfo>("MG_Impl/GLImpl", caller,
+                                             "This operation is not allowed on the default texture (zero is "
+                                             "bound to the target)."));
+            return false;
+        }
+        return true;
+    }
+
     Bool ValidateTextureTargetUniformity(SharedPtr<MG_State::GLState::ITextureObject> textureObject,
                                          TextureTarget target) {
         if (!textureObject) return true; // should be created later