Merge branch 'master' into sme1_gemm_support

Author: vgundlur

.clang-format

@@ -1,34 +1 @@
-AllowShortFunctionsOnASingleLine: Inline
-PackConstructorInitializers: Never
-ColumnLimit: 120
-AlignAfterOpenBracket: AlwaysBreak
-BinPackParameters: false
-AllowAllParametersOfDeclarationOnNextLine: true
-BreakBeforeBraces: Stroustrup
-SpaceAfterCStyleCast: true
-PointerAlignment: Left
-ForEachMacros: ['XNN_UNPREDICTABLE', 'XNN_LIKELY', 'XNN_UNLIKELY']
-IfMacros: ['IF']
-IndentCaseLabels: true
-ContinuationIndentWidth: 2
-SpaceBeforeParens: Custom
-SpaceBeforeParensOptions:
-  AfterControlStatements: true
-  AfterIfMacros: true
-  AfterForeachMacros: false
-SpacesBeforeTrailingComments: 2
-IncludeBlocks: Regroup
-IncludeCategories:
-  - Regex: '<xnnpack[./][[:alnum:].-]+>' # match XNNPack includes first
-    Priority: 5
-  - Regex: 'benchmark.h' # includes used in benchmarks
-    Priority: 3
-  - Regex: 'bench/' # includes used in benchmarks
-    Priority: 3
-  - Regex: 'gtest.h' # includes used in tests
-    Priority: 3
-  - Regex: 'gmock.h' # includes used in tests
-    Priority: 3
-  - Regex: '<[[:alnum:].]+>' # system headers
-    Priority: 2 # lower priority to keep it sorted first before XNNPack includes
-MaxEmptyLinesToKeep: 2 # used to separate includes from functions
+BasedOnStyle: Google

.github/workflows/build.yml

@@ -20,12 +20,15 @@ concurrency:
 
 jobs:
   cmake-linux-x86_64:
-    runs-on: ubuntu-22.04-8core
+    runs-on: 'ubuntu-24.04-8core'
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install ninja
         run: sudo apt install ninja-build
       - name: Setup ccache
@@ -41,13 +44,17 @@ jobs:
         if: ${{ inputs.run-tests }}
         run: ctest --output-on-failure --parallel $(nproc)
         working-directory: ${{ github.workspace }}/build/local
+
   cmake-linux-aarch64:
     runs-on: arm-ubuntu-arm-22.04-8core
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install ninja
         run: sudo apt install ninja-build
       - name: Setup ccache
@@ -63,13 +70,17 @@ jobs:
         if: ${{ inputs.run-tests }}
         run: ctest --output-on-failure --parallel $(nproc)
         working-directory: ${{ github.workspace }}/build/local
+
   cmake-linux-armhf:
     runs-on: arm-ubuntu-arm-22.04-8core
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install ninja
         run: sudo apt install ninja-build
       - name: Install armhf cross-toolchain
@@ -91,13 +102,17 @@ jobs:
           LD_LIBRARY_PATH: /usr/arm-linux-gnueabihf/lib/:${{ env.LD_LIBRARY_PATH }}
         run: ctest --output-on-failure --parallel $(nproc)
         working-directory: ${{ github.workspace }}/build/linux/armhf
+
   cmake-linux-riscv64:
-    runs-on: ubuntu-22.04-16core
+    runs-on: 'ubuntu-24.04-16core'
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install ninja
         run: sudo apt install ninja-build
       - name: Install riscv64 cross-toolchain
@@ -117,39 +132,62 @@ jobs:
         if: ${{ inputs.run-tests }}
         run: ctest --output-on-failure --parallel $(nproc)
         working-directory: ${{ github.workspace }}/build/linux/riscv64
+
   cmake-linux-riscv64-rvv:
-    runs-on: ubuntu-22.04-16core
+    runs-on: 'ubuntu-24.04-16core'
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install ninja
         run: sudo apt install ninja-build
       - name: Download and Install riscv64 cross-toolchain tarball
         run: |
-          wget -nv https://github.com/riscv-collab/riscv-gnu-toolchain/releases/download/2024.09.03/riscv64-glibc-ubuntu-22.04-llvm-nightly-2024.09.03-nightly.tar.gz
-          tar xzf riscv64-glibc-ubuntu-22.04-llvm-nightly-2024.09.03-nightly.tar.gz
+          wget -nv https://github.com/riscv-collab/riscv-gnu-toolchain/releases/download/2025.01.20/riscv64-glibc-ubuntu-24.04-llvm-nightly-2025.01.20-nightly.tar.xz
+          tar xJf riscv64-glibc-ubuntu-24.04-llvm-nightly-2025.01.20-nightly.tar.xz
         working-directory: ${{ github.workspace }}
       - name: Setup ccache
         uses: hendrikmuhs/[email protected]
         with:
           key: ${{ github.job }}
           max-size: "500M"
           save: ${{ inputs.update-caches }}
+      - name: Check for cached qemu build
+        id: cached-qemu
+        uses: actions/cache@v4
+        with:
+          path: ${{ github.workspace }}/bin/qemu-riscv64
+          key: ${{ runner.os }}-qemu-riscv64
+      - name: Download and build qemu-riscv64
+        if: steps.cached-qemu.outputs.cache-hit != 'true'
+        run: |
+          sudo apt install libglib2.0-dev
+          mkdir bin
+          wget -nv https://download.qemu.org/qemu-9.2.2.tar.xz
+          tar xJf qemu-9.2.2.tar.xz
+          cd qemu-9.2.2
+          ./configure --target-list=riscv64-linux-user --static
+          make -j $(nproc)
+          cp build/qemu-riscv64 ../bin/
+        working-directory: ${{ github.workspace }}
       - name: Configure and build
         run: |
           scripts/build-linux-riscv64.sh \
             -DCMAKE_BUILD_TYPE=Release \
             -DXNNPACK_ENABLE_RISCV_VECTOR=ON \
             -DUSE_GNU_SOURCE=ON \
             -DRISCV_TOOLCHAIN_ROOT=${{ github.workspace }}/riscv/ \
-            -DRISCV_QEMU_ROOT=${{ github.workspace }}/riscv/
+            -DRISCV_QEMU_ROOT=${{ github.workspace }}
         working-directory: ${{ github.workspace }}
       - name: Run tests (no operator tests)
         if: ${{ inputs.run-tests }}
-        run: ctest --output-on-failure -LE operator --parallel $(nproc)
+        run: ctest --output-on-failure --label-exclude operator --parallel $(nproc)
         working-directory: ${{ github.workspace }}/build/linux/riscv64
+
   cmake-windows-arm64:
     runs-on: windows-2022-32core
     timeout-minutes: 60
@@ -174,6 +212,7 @@ jobs:
         env:
           CFLAGS: "/UNDEBUG"
           CXXFLAGS: "/UNDEBUG"
+
   cmake-windows-x64:
     runs-on: windows-2022-32core
     timeout-minutes: 60
@@ -202,6 +241,7 @@ jobs:
         if: ${{ inputs.run-tests }}
         run: ctest -C Release --output-on-failure --parallel $NUMBER_OF_PROCESSORS
         working-directory: ${{ github.workspace }}/build/windows/x64
+
   cmake-windows-x86:
     runs-on: windows-2022-32core
     timeout-minutes: 60
@@ -230,6 +270,7 @@ jobs:
         if: ${{ inputs.run-tests }}
         run: ctest -C Release --output-on-failure --parallel $NUMBER_OF_PROCESSORS
         working-directory: ${{ github.workspace }}/build/windows/x86
+
   cmake-macos-arm64:
     runs-on: macos-latest
     timeout-minutes: 60
@@ -262,16 +303,20 @@ jobs:
             build/macos/arm64 \
             -j$((2*$(sysctl -n hw.ncpu)))
         working-directory: ${{ github.workspace }}
+
   cmake-android:
     strategy:
       matrix:
         arch: [arm64, armv7, x86]
-    runs-on: ubuntu-22.04-8core
+    runs-on: 'ubuntu-24.04-8core'
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install ninja
         run: sudo apt install ninja-build
       - name: Setup Android NDK
@@ -300,6 +345,7 @@ jobs:
         working-directory: ${{ github.workspace }}
         env:
           ANDROID_NDK: ${{ steps.setup-ndk.outputs.ndk-path }}
+
   cmake-ios-arm64:
     runs-on: macos-latest
     timeout-minutes: 60
@@ -314,6 +360,7 @@ jobs:
       - name: Build with Xcode
         run: cmake --build build/ios/arm64 --parallel $(sysctl -n hw.ncpu) -- -quiet
         working-directory: ${{ github.workspace }}
+
   cmake-ios-x86_64:
     runs-on: macos-latest
     timeout-minutes: 60
@@ -328,13 +375,17 @@ jobs:
       - name: Build with Xcode
         run: cmake --build build/ios/x86_64 --parallel $(sysctl -n hw.ncpu) -- -sdk iphonesimulator -quiet
         working-directory: ${{ github.workspace }}
+
   bazel-linux-x86_64-clang-18:
-    runs-on: ubuntu-22.04-8core
+    runs-on: 'ubuntu-24.04-8core'
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install clang-18
         working-directory: ${{ github.workspace }}
         run: |
@@ -374,8 +425,9 @@ jobs:
         with:
           path: "/home/runner/.cache/bazel"
           key: ${{ github.job }}-${{ github.sha }}
+
   bazel-linux-x86_64-gcc-9:
-    runs-on: ubuntu-22.04-8core
+    runs-on: 'ubuntu-24.04-8core'
     timeout-minutes: 60
     env:
       CC: gcc-9
@@ -384,7 +436,10 @@ jobs:
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install gcc-9
         working-directory: ${{ github.workspace }}
         run: |
@@ -416,13 +471,17 @@ jobs:
         with:
           path: "/home/runner/.cache/bazel"
           key: ${{ github.job }}-${{ github.sha }}
+
   bazel-linux-aarch64-clang18:
     runs-on: arm-ubuntu-arm-22.04-4core
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
       - name: Update apt
-        run: sudo apt update
+        run: |
+          echo 'set man-db/auto-update false' | sudo debconf-communicate >/dev/null
+          sudo dpkg-reconfigure man-db
+          sudo apt update
       - name: Install clang-18
         working-directory: ${{ github.workspace }}
         run: |
@@ -454,8 +513,8 @@ jobs:
             --test_output=all \
             --test_filter=*QP8* \
             test:qp8_f32_qc4w_gemm_minmax_test \
-            test:fully_connected_nc_test \
-            test:fully_connected_test
+            test/operators:fully_connected_nc_test \
+            test/subgraph:fully_connected_test
         working-directory: ${{ github.workspace }}
       - name: Compress disk cache
         # Bazel's `--disk-cache` currently grows without bounds, so we remove files
@@ -468,6 +527,7 @@ jobs:
         with:
           path: "/home/runner/.cache/bazel"
           key: ${{ github.job }}-${{ github.sha }}
+
   bazel-linux-aarch64-gcc13:
     # This ensures we have solid test coverage for _Float16
     runs-on: arm-ubuntu-arm-22.04-4core
@@ -509,8 +569,8 @@ jobs:
             --test_output=all \
             --test_filter=*QP8* \
             test:qp8_f32_qc4w_gemm_minmax_test \
-            test:fully_connected_nc_test \
-            test:fully_connected_test
+            test/operators:fully_connected_nc_test \
+            test/subgraph:fully_connected_test
         working-directory: ${{ github.workspace }}
       - name: Compress disk cache
         # Bazel's `--disk-cache` currently grows without bounds, so we remove files

.github/workflows/sde-tests-linux-windows.yml

@@ -16,14 +16,15 @@ jobs:
     strategy:
       matrix:
         arch: [windows-x86, windows-x64, linux-x86_64]
-        sde-processor: [wsm, snb, ivb, hsw, bdw, skl, cnl, icl, skx, clx, cpx, icx, tgl, adl, mtl, rpl, spr, emr, gnr, gnr256, dmr, srf, arl, lnl, cwf, future]
+        sde-processor: [slm, wsm, snb, ivb, hsw, bdw, skl, cnl, icl, skx, clx, cpx, icx, tgl, adl, mtl, rpl, spr, emr, gnr, gnr256, dmr, srf, arl, lnl, cwf]
     runs-on: ${{ matrix.arch == 'linux-x86_64' && 'ubuntu-latest' || 'windows-latest' }}
     needs: build
     continue-on-error: true
 
     name: >
       (${{ matrix.arch }}, ${{ matrix.sde-processor }})
       - ${{ fromJson('{
+        "slm": {"main": "Silvermont Processor - Supports SSE41"},
         "wsm": {"main": "Westmere Processor - Supports SSE41"},
         "snb": {"main": "Sandy Bridge Processor - Supports SSE41, AVX"},
         "ivb": {"main": "Ivy Bridge Processor - Supports SSE41, AVX, F16C"},
@@ -48,8 +49,7 @@ jobs:
         "srf": {"main": "Sierra Forest Processor - Supports SSE41, AVX, F16C, FMA3, AVX2, AVXVNNI, AVXVNNIINT8"},
         "arl": {"main": "Arrow Lake Processor - Supports SSE41, AVX, F16C, FMA3, AVX2, AVXVNNI, AVXVNNIINT8"},
         "lnl": {"main": "Lunar Lake Processor - Supports SSE41, AVX, F16C, FMA3, AVX2, AVXVNNI, AVXVNNIINT8"},
-        "cwf": {"main": "Clearwater Forest Processor - Supports SSE41, AVX, F16C, FMA3, AVX2, AVXVNNI, AVXVNNIINT8"},
-        "future": {"main": "Future Chip Processor - Supports SSE41, AVX, F16C, FMA3, AVX2, AVX512F, AVX512FP16, AVX512SKX, AVX256SKX, AVX512VBMI, AVX512VNNI, AVX512VNNIGFNI, AVX256VNNI, AVX256VNNIGFNI, AVX512BF16, AVXVNNI, AVXVNNIINT8"}
+        "cwf": {"main": "Clearwater Forest Processor - Supports SSE41, AVX, F16C, FMA3, AVX2, AVXVNNI, AVXVNNIINT8"}
          }')[matrix.sde-processor].main }}
 
     steps:

.gitignore

@@ -34,3 +34,4 @@ ehthumbs.db
 Thumbs.db
 *.swp
 .vscode
+.vs