Merge branch 'sycl-develop' into zt/perf

Author: taozha2

.github/actions/install-dpcpp/action.yml

@@ -36,32 +36,66 @@ runs:
           URL=https://github.com/intel/llvm/releases/download/${{ inputs.DPCPP_VERSION }}/sycl_linux.tar.gz;
         else
           echo "Will use latest DPCPP version"
-          latest=$(curl -sS https://api.github.com/repos/intel/llvm/releases | jq -r '[.[].tag_name|select(match("nightly-[0-9]{4}-[0-9]{2}-[0-9]{2}"))][0]')
+          # Use grep and sed instead of jq to avoid additional dependencies
+          latest=$(curl -sS https://api.github.com/repos/intel/llvm/releases | \
+            grep '"tag_name":' | \
+            grep 'nightly-[0-9]\{4\}-[0-9]\{2\}-[0-9]\{2\}' | \
+            head -1 | \
+            sed 's/.*"tag_name": *"\([^"]*\)".*/\1/')
           URL=https://github.com/intel/llvm/releases/download/${latest}/sycl_linux.tar.gz;
         fi
         echo "Downloading DPCPP from ${URL}"
         sudo wget -q $URL
         sudo tar -xf sycl_linux.tar.gz
         sudo rm sycl_linux.tar.gz
         popd
+        
+        # Detect system C++ compiler and standard library automatically
+        echo "=== Detecting system C++ environment ==="
+        SYSTEM_GCC_VERSION=$(gcc --version | head -1 | grep -o '[0-9]\+\.[0-9]\+' | head -1 | cut -d. -f1)
+        SYSTEM_CXX_INCLUDE_DIR="/usr/include/c++/${SYSTEM_GCC_VERSION}"
+        SYSTEM_CXX_ARCH_INCLUDE_DIR="/usr/include/x86_64-linux-gnu/c++/${SYSTEM_GCC_VERSION}"
+        SYSTEM_GCC_LIB_DIR="/usr/lib/gcc/x86_64-linux-gnu/${SYSTEM_GCC_VERSION}"
+        
+        echo "Detected GCC version: ${SYSTEM_GCC_VERSION}"
+        echo "C++ include directory: ${SYSTEM_CXX_INCLUDE_DIR}"
+        echo "C++ arch include directory: ${SYSTEM_CXX_ARCH_INCLUDE_DIR}"
+        echo "GCC library directory: ${SYSTEM_GCC_LIB_DIR}"
+        
+        # Verify directories exist, fallback to available versions if needed
+        if [[ ! -d "$SYSTEM_CXX_INCLUDE_DIR" ]]; then
+          echo "Primary C++ include dir not found, searching for alternatives..."
+          SYSTEM_CXX_INCLUDE_DIR=$(find /usr/include/c++/ -maxdepth 1 -type d -name '[0-9]*' | sort -V | tail -1)
+          if [[ -n "$SYSTEM_CXX_INCLUDE_DIR" ]]; then
+            DETECTED_VERSION=$(basename "$SYSTEM_CXX_INCLUDE_DIR")
+            SYSTEM_CXX_ARCH_INCLUDE_DIR="/usr/include/x86_64-linux-gnu/c++/${DETECTED_VERSION}"
+            SYSTEM_GCC_LIB_DIR="/usr/lib/gcc/x86_64-linux-gnu/${DETECTED_VERSION}"
+            echo "Using fallback C++ version: ${DETECTED_VERSION}"
+          fi
+        fi
+        
         cat <<EOF | tee setvars.sh
-        export PATH=$DPCPP_PATH/bin:$PATH
-        export C_INCLUDE_PATH=$DPCPP_PATH/include:$C_INCLUDE_PATH
-        export CPLUS_INCLUDE_PATH=$DPCPP_PATH/include:$CPLUS_INCLUDE_PATH
-        export LD_LIBRARY_PATH=$DPCPP_PATH/lib:$LD_LIBRARY_PATH
+        export PATH=$DPCPP_PATH/bin:/bin:/usr/bin:/usr/local/bin:\$PATH
+        export C_INCLUDE_PATH=$DPCPP_PATH/include:${SYSTEM_CXX_INCLUDE_DIR}:${SYSTEM_CXX_ARCH_INCLUDE_DIR}:/usr/include/x86_64-linux-gnu:/usr/include:\$C_INCLUDE_PATH
+        export CPLUS_INCLUDE_PATH=$DPCPP_PATH/include:${SYSTEM_CXX_INCLUDE_DIR}:${SYSTEM_CXX_ARCH_INCLUDE_DIR}:/usr/include/x86_64-linux-gnu:/usr/include:\$CPLUS_INCLUDE_PATH
+        export LD_LIBRARY_PATH=$DPCPP_PATH/lib:/usr/lib/x86_64-linux-gnu:${SYSTEM_GCC_LIB_DIR}:/usr/lib:\$LD_LIBRARY_PATH
+        export LIBRARY_PATH=/usr/lib/x86_64-linux-gnu:${SYSTEM_GCC_LIB_DIR}:/usr/lib:\$LIBRARY_PATH
         export CC=$DPCPP_PATH/bin/clang
         export CXX=$DPCPP_PATH/bin/clang++
+        # Configure DPC++ to use system C++ standard library with proper linking
+        export CXXFLAGS="-stdlib=libstdc++ -I${SYSTEM_CXX_INCLUDE_DIR} -I${SYSTEM_CXX_ARCH_INCLUDE_DIR} \$CXXFLAGS"
+        export LDFLAGS="-L/usr/lib/x86_64-linux-gnu -L${SYSTEM_GCC_LIB_DIR} -L/usr/lib \$LDFLAGS"
         EOF
     - name: Install DPCPP
       if: inputs.DPCPP_RELEASE == 'RELEASE'
       shell: bash
       run: |
         shopt -s expand_aliases
         which sudo || alias sudo=""
-        wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-        sudo apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-        rm GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-        echo "deb https://apt.repos.intel.com/oneapi all main" | sudo tee /etc/apt/sources.list.d/oneAPI.list
+        wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB | \
+          gpg --dearmor | sudo tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null
+        echo "deb [signed-by=/usr/share/keyrings/oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main" | \
+          sudo tee /etc/apt/sources.list.d/oneAPI.list
         sudo apt update
         sudo apt install -y intel-oneapi-runtime-libs intel-oneapi-compiler-dpcpp-cpp
         . /opt/intel/oneapi/setvars.sh

.github/actions/install-intel-graphics/action.yml

@@ -21,7 +21,10 @@ runs:
           sudo add-apt-repository ppa:kobuk-team/intel-graphics
           sudo apt update
         else
-          sudo apt update
+          . /etc/os-release
+          wget https://repositories.intel.com/gpu/ubuntu/dists/${VERSION_CODENAME}/intel-gpu-ubuntu-${VERSION_CODENAME}.run
+          chmod +x intel-gpu-ubuntu-${VERSION_CODENAME}.run
+          sudo ./intel-gpu-ubuntu-${VERSION_CODENAME}.run
           sudo apt install -y \
             intel-media-va-driver-non-free libmfx-gen1 libvpl2 \
             libegl-mesa0 libegl1-mesa-dev libgl1-mesa-dev  \
@@ -39,7 +42,7 @@ runs:
         shopt -s expand_aliases
         which sudo || alias sudo=""
         if [[ "${{ inputs.GPU }}" == "BMG" ]]; then
-          sudo rm /etc/apt/sources.list.d/ppa_kobuk_team_intel_graphics_oracular.list
+          sudo rm -f /etc/apt/sources.list.d/ppa_kobuk_team_intel_graphics_oracular.list
           sudo add-apt-repository ppa:kobuk-team/intel-graphics-staging
           sudo apt update
           sudo apt-get install -y libze-intel-gpu1 libze-dev intel-metrics-discovery \

.github/workflows/intel_test.yml

@@ -28,36 +28,36 @@ jobs:
             gpu: BMG
             intel_graphics: STAGING
             sycl_target: intel_gpu_bmg_g21
-            runner: cp-b580-gpu
+            runner: bmg108629-01
           - compiler: RELEASE
             gpu: BMG
             intel_graphics: STAGING
             sycl_target: intel_gpu_bmg_g21
-            runner: cp-b580-gpu
-          - compiler: RELEASE
+            runner: bmg108629-01
+          - compiler: NIGHTLY
             gpu: BMG
             intel_graphics: ROLLING
             sycl_target: intel_gpu_bmg_g21
-            runner: cp-b580-gpu
-          - compiler: NIGHTLY
+            runner: bmg108629-01
+          - compiler: RELEASE
             gpu: BMG
             intel_graphics: ROLLING
             sycl_target: intel_gpu_bmg_g21
-            runner: cp-b580-gpu
-          - compiler: NIGHTLY
+            runner: bmg108629-01
+          - compiler: RELEASE
             gpu: PVC
             intel_graphics: ROLLING
             sycl_target: intel_gpu_pvc
-            runner: cp-gpumax-1100-gpu
-          - compiler: RELEASE
+            runner: pvc146162-01
+          - compiler: NIGHTLY
             gpu: PVC
             intel_graphics: ROLLING
             sycl_target: intel_gpu_pvc
-            runner: cp-gpumax-1100-gpu
+            runner: pvc146162-01
 
     name: Run Intel ${{ matrix.compiler }} tests on ${{ matrix.gpu }} with intel-graphics ${{ matrix.intel_graphics }}
     runs-on: ${{ matrix.runner }}
-    timeout-minutes: 30
+    timeout-minutes: 120
 
     steps:
       - name: Checkout repository
@@ -79,9 +79,17 @@ jobs:
       - name: Setup virtual environment
         shell: bash
         run: |
+          # Install cmake and ninja if not already available
+          if ! command -v cmake &> /dev/null || ! command -v ninja &> /dev/null; then
+            echo "Installing cmake and/or ninja..."
+            sudo apt update
+            sudo apt install -y cmake ninja-build
+          else
+            echo "cmake and ninja already available"
+          fi
           . setvars.sh
           export IGC_ExtraOCLOptions="-cl-intel-256-GRF-per-thread"
-          export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file -gline-tables-only" 
+          export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file -gline-tables-only"
           export ONEAPI_DEVICE_SELECTOR=level_zero:gpu
           export IGC_VectorAliasBBThreshold=100000000000
           # Persist environment variables to following steps 
@@ -96,11 +104,16 @@ jobs:
             -DCUTLASS_ENABLE_SYCL=ON \
             -DDPCPP_SYCL_TARGET=${{ matrix.sycl_target }} \
             -DCUTLASS_SYCL_RUNNING_CI=ON
-          cmake --build .
+          # Use reduced parallelism for BMG runners, full for PVC
+          if [[ "${{ matrix.gpu }}" == "BMG" ]]; then
+            cmake --build . -j 2
+          else
+            cmake --build . 
+          fi
       - name: Unit test
         shell: bash
         run: |
-          cmake --build . --target test_unit
+          cmake --build . --target test_unit -j 8
       - name: Examples
         shell: bash
         run: |
@@ -109,3 +122,30 @@ jobs:
         shell: bash
         run: |
           cmake --build . --target cutlass_benchmarks
+      - name: Cleanup DPC++
+        if: always()
+        shell: bash
+        run: |
+          echo "Cleaning up DPC++ installation..."
+          # Remove DPCPP directory if it exists
+          DPCPP_PATH="${{ inputs.DPCPP_PATH || '~/dpcpp' }}"
+          DPCPP_PATH=$(eval echo $DPCPP_PATH)  # Expand ~ to home directory
+          if [ -d "$DPCPP_PATH" ]; then
+            echo "Removing DPCPP directory: $DPCPP_PATH"
+            sudo rm -rf "$DPCPP_PATH"
+          fi
+          # For RELEASE installs, remove OneAPI packages
+          if [[ "${{ matrix.compiler }}" == "RELEASE" ]]; then
+            echo "Removing OneAPI packages..."
+            sudo apt remove -y intel-oneapi-runtime-libs intel-oneapi-compiler-dpcpp-cpp || true
+            sudo rm -f /etc/apt/sources.list.d/oneAPI.list
+            sudo rm -f /usr/share/keyrings/oneapi-archive-keyring.gpg
+          fi
+          # Clean up environment files
+          rm -f setvars.sh
+          # Clean up build artifacts
+          rm -rf build/ || true
+          # Reset environment variables that might interfere
+          unset CC CXX CPLUS_INCLUDE_PATH C_INCLUDE_PATH LD_LIBRARY_PATH
+          unset IGC_ExtraOCLOptions SYCL_PROGRAM_COMPILE_OPTIONS ONEAPI_DEVICE_SELECTOR IGC_VectorAliasBBThreshold
+          echo "DPC++ cleanup completed"