Merge branch 'sycl' into simplify-device-global

Author: zhaomaosu

.github/docker/install_dpcpp.sh

@@ -16,5 +16,5 @@ if [ "${SKIP_DPCPP_BUILD}" ]; then
 fi
 
 mkdir -p ${DPCPP_PATH}/dpcpp_compiler
-wget -O ${DPCPP_PATH}/dpcpp_compiler.tar.gz https://github.com/intel/llvm/releases/download/nightly-2024-01-29/sycl_linux.tar.gz
+wget -O ${DPCPP_PATH}/dpcpp_compiler.tar.gz https://github.com/intel/llvm/releases/download/nightly-2024-09-27/sycl_linux.tar.gz
 tar -xvf ${DPCPP_PATH}/dpcpp_compiler.tar.gz -C ${DPCPP_PATH}/dpcpp_compiler

.github/workflows/benchmarks_compute.yml

@@ -151,13 +151,32 @@ jobs:
     - name: Build SYCL
       run: cmake --build ${{github.workspace}}/sycl_build -j
 
-    - name: Set oneAPI Device Selector
-      run: |
-        echo "ONEAPI_DEVICE_SELECTOR=${{ matrix.adapter.str_name }}:${{ matrix.adapter.unit }}" >> $GITHUB_ENV
+    - name: Configure UR
+      run: >
+        cmake -DCMAKE_BUILD_TYPE=Release
+        -S${{github.workspace}}/ur-repo
+        -B${{github.workspace}}/ur_build
+        -DCMAKE_INSTALL_PREFIX=${{github.workspace}}/ur_install
+        -DUR_BUILD_TESTS=OFF
+        -DUR_BUILD_ADAPTER_L0=ON
+        -DUR_BUILD_ADAPTER_L0_V2=ON
+        -DUMF_DISABLE_HWLOC=ON
+
+    - name: Build UR
+      run: cmake --build ${{github.workspace}}/ur_build -j $(nproc)
+
+    - name: Install UR
+      run: cmake --install ${{github.workspace}}/ur_build
 
     - name: Run benchmarks
       id: benchmarks
-      run: numactl -N 0 ${{ github.workspace }}/ur-repo/scripts/benchmarks/main.py ~/bench_workdir ${{github.workspace}}/sycl_build ${{github.workspace}}/ur-repo ${{ matrix.adapter.str_name }} ${{ inputs.bench_script_params }}
+      run: >
+        numactl -N 0 ${{ github.workspace }}/ur-repo/scripts/benchmarks/main.py 
+        ~/bench_workdir 
+        ${{github.workspace}}/sycl_build
+        ${{github.workspace}}/ur_install
+        ${{ matrix.adapter.str_name }}
+        ${{ inputs.bench_script_params }}
 
     - name: Add comment to PR
       uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1

.github/workflows/build-hw-reusable.yml

@@ -18,6 +18,10 @@ on:
         required: false
         type: string
         default: OFF
+      static_adapter:
+        required: false
+        type: string
+        default: OFF
 
 permissions:
   contents: read
@@ -36,7 +40,7 @@ jobs:
     strategy:
       matrix:
         adapter: [
-          {name: "${{inputs.adapter_name}}", platform: "${{inputs.platform}}", static_Loader: "${{inputs.static_loader}}"},
+          {name: "${{inputs.adapter_name}}", platform: "${{inputs.platform}}", static_Loader: "${{inputs.static_loader}}", static_adapter: "${{inputs.static_loader}}"},
         ]
         build_type: [Debug, Release]
         compiler: [{c: gcc, cxx: g++}, {c: clang, cxx: clang++}]
@@ -49,6 +53,10 @@ jobs:
            build_type: Release
          - adapter: {static_Loader: ON}
            compiler: {c: clang, cxx: clang++}
+         - adapter: {static_adapter: ON}
+           build_type: Release
+         - adapter: {static_adapter: ON}
+           compiler: {c: clang, cxx: clang++}
 
     runs-on: ${{inputs.runner_name}}
 
@@ -76,15 +84,21 @@ jobs:
         -DUR_BUILD_TESTS=ON
         -DUR_BUILD_ADAPTER_${{matrix.adapter.name}}=ON
         -DUR_STATIC_LOADER=${{matrix.adapter.static_Loader}}
+        -DUR_STATIC_ADAPTER_${{matrix.adapter.name}}=${{matrix.adapter.static_adapter}}
         -DUR_DPCXX=${{github.workspace}}/dpcpp_compiler/bin/clang++
         -DUR_SYCL_LIBRARY_DIR=${{github.workspace}}/dpcpp_compiler/lib
+        -DCMAKE_INSTALL_PREFIX=${{github.workspace}}/install
         ${{ matrix.adapter.name == 'HIP' && '-DUR_CONFORMANCE_AMD_ARCH=gfx1030' || '' }}
         ${{ matrix.adapter.name == 'HIP' && '-DUR_HIP_PLATFORM=AMD' || '' }}
 
     - name: Build
       # This is so that device binaries can find the sycl runtime library
       run: cmake --build ${{github.workspace}}/build -j $(nproc)
 
+    - name: Install
+      # This is to check that install command does not fail
+      run: cmake --install ${{github.workspace}}/build
+
     - name: Test adapter specific
       working-directory: ${{github.workspace}}/build
       run: ctest -C ${{matrix.build_type}} --output-on-failure -L "adapter-specific" --timeout 180

.github/workflows/cmake.yml

@@ -20,7 +20,7 @@ jobs:
         compiler: [{c: gcc, cxx: g++}]
         libbacktrace: ['-DVAL_USE_LIBBACKTRACE_BACKTRACE=OFF']
         pool_tracking: ['-DUMF_ENABLE_POOL_TRACKING=ON', '-DUMF_ENABLE_POOL_TRACKING=OFF']
-        latency_tracking: ['-DUMF_ENABLE_LATENCY_TRACKING=OFF']
+        latency_tracking: ['-DUR_ENABLE_LATENCY_HISTOGRAM=OFF']
         include:
           - os: 'ubuntu-22.04'
             build_type: Release
@@ -40,7 +40,7 @@ jobs:
           - os: 'ubuntu-22.04'
             build_type: Release
             compiler: {c: clang, cxx: clang++}
-            latency_tracking: '-DUMF_ENABLE_LATENCY_TRACKING=ON'
+            latency_tracking: '-DUR_ENABLE_LATENCY_HISTOGRAM=ON'
     runs-on: ${{ (matrix.os == 'ubuntu-22.04' && github.repository_owner == 'oneapi-src') && 'intel-ubuntu-22.04' || matrix.os }}
 
     steps:
@@ -49,7 +49,7 @@ jobs:
     - name: Install apt packages
       run: |
         sudo apt-get update
-        sudo apt-get install -y doxygen ${{matrix.compiler.c}}
+        sudo apt-get install -y ${{matrix.compiler.c}}
 
     - name: Install libhwloc
       run: .github/scripts/install_hwloc.sh
@@ -62,9 +62,6 @@ jobs:
       run: |
         sudo apt-get install -y ${{matrix.compiler.cxx}}
 
-    - name: Install pip packages
-      run: pip install -r third_party/requirements.txt
-
     - name: Install libbacktrace
       if: matrix.libbacktrace == '-DVAL_USE_LIBBACKTRACE_BACKTRACE=ON'
       run: |
@@ -79,7 +76,7 @@ jobs:
       if: matrix.os == 'ubuntu-22.04'
       run: |
         sudo apt install libncurses5
-        wget -O ${{github.workspace}}/dpcpp_compiler.tar.gz https://github.com/intel/llvm/releases/download/nightly-2024-01-29/sycl_linux.tar.gz
+        wget -O ${{github.workspace}}/dpcpp_compiler.tar.gz https://github.com/intel/llvm/releases/download/nightly-2024-09-27/sycl_linux.tar.gz
         mkdir -p ${{github.workspace}}/dpcpp_compiler
         tar -xvf ${{github.workspace}}/dpcpp_compiler.tar.gz -C ${{github.workspace}}/dpcpp_compiler
 
@@ -93,7 +90,7 @@ jobs:
         -DUR_ENABLE_TRACING=ON
         -DCMAKE_BUILD_TYPE=${{matrix.build_type}}
         -DUR_BUILD_TESTS=ON
-        -DUR_FORMAT_CPP_STYLE=ON
+        -DUR_FORMAT_CPP_STYLE=OFF
         -DUR_DPCXX=${{github.workspace}}/dpcpp_compiler/bin/clang++
         ${{matrix.libbacktrace}}
         ${{matrix.pool_tracking}}
@@ -109,18 +106,11 @@ jobs:
         -DUR_ENABLE_TRACING=ON
         -DCMAKE_BUILD_TYPE=${{matrix.build_type}}
         -DUR_BUILD_TESTS=ON
-        -DUR_FORMAT_CPP_STYLE=ON
+        -DUR_FORMAT_CPP_STYLE=OFF
         ${{matrix.libbacktrace}}
         ${{matrix.pool_tracking}}
         ${{matrix.latency_tracking}}
 
-    - name: Generate source from spec, check for uncommitted diff
-      if: matrix.os == 'ubuntu-22.04'
-      run: cmake --build ${{github.workspace}}/build --target check-generated
-
-    - name: Verify that each source file contains a license
-      run: cmake --build ${{github.workspace}}/build --target verify-licenses
-
     - name: Build
       run: cmake --build ${{github.workspace}}/build -j $(nproc)
 
@@ -155,6 +145,7 @@ jobs:
       adapter_name: L0
       runner_name: L0
       static_loader: ON
+      static_adapter: ON
 
   opencl:
     name: OpenCL
@@ -216,7 +207,8 @@ jobs:
         os: ['windows-2019', 'windows-2022']
         adapter: [
           {name: None, var: ''}, {name: L0, var: '-DUR_BUILD_ADAPTER_L0=ON'},
-          {name: None, var: ''}, {name: L0_V2, var: '-DUR_BUILD_ADAPTER_L0_V2=ON'}
+          {name: None, var: ''}, {name: L0_V2, var: '-DUR_BUILD_ADAPTER_L0_V2=ON'},
+          {name: L0, var: '-DUR_BUILD_ADAPTER_L0=ON -DUR_STATIC_ADAPTER_L0=ON'}
         ]
 
         # TODO: building level zero loader on windows-2019 and clang-cl is currently broken
@@ -225,16 +217,25 @@ jobs:
            adapter: {name: L0, var: '-DUR_BUILD_ADAPTER_L0=ON'}
          - os: 'windows-2019'
            adapter: {name: L0_V2, var: '-DUR_BUILD_ADAPTER_L0_V2=ON'}
+         - os: 'windows-2019'
+           adapter: {name: L0, var: '-DUR_BUILD_ADAPTER_L0=ON -DUR_STATIC_ADAPTER_L0=ON'}
          - adapter: {name: L0, var: '-DUR_BUILD_ADAPTER_L0=ON'}
            compiler: {c: clang-cl, cxx: clang-cl}
          - adapter: {name: L0_V2, var: '-DUR_BUILD_ADAPTER_L0_V2=ON'}
            compiler: {c: clang-cl, cxx: clang-cl}
+         - adapter: {name: L0, var: '-DUR_BUILD_ADAPTER_L0=ON -DUR_STATIC_ADAPTER_L0=ON'}
+           compiler: {c: clang-cl, cxx: clang-cl}
 
         build_type: [Debug, Release]
         compiler: [{c: cl, cxx: cl}, {c: clang-cl, cxx: clang-cl}]
         include:
           - compiler: {c: clang-cl, cxx: clang-cl}
             toolset: "-T ClangCL"
+          - os: 'windows-2022'
+            adapter: {name: L0, var: '-DUR_BUILD_ADAPTER_L0=ON -DUR_STATIC_ADAPTER_L0=ON'}
+            build_type: 'Release'
+            compiler: {c: cl, cxx: cl}
+
     runs-on: ${{matrix.os}}
 
     steps:
@@ -274,11 +275,6 @@ jobs:
         -DUR_FORMAT_CPP_STYLE=ON
         ${{matrix.adapter.var}}
 
-    # TODO: re-enable when check-generated is fixed for windows runners see #888
-    # - name: Generate source from spec, check for uncommitted diff
-    #   if: matrix.os == 'windows-2022'
-    #   run: cmake --build ${{github.workspace}}/build --target check-generated --config ${{matrix.build_type}}
-
     - name: Build all
       run: cmake --build ${{github.workspace}}/build --config ${{matrix.build_type}} -j $Env:NUMBER_OF_PROCESSORS
 

.github/workflows/e2e_core.yml

@@ -169,10 +169,6 @@ jobs:
         -DCMAKE_CXX_COMPILER="$(which clang++)"
         -DLLVM_LIT="${{github.workspace}}/sycl-repo/llvm/utils/lit/lit.py"
 
-    - name: Set LIT_XFAIL_NOT
-      if: inputs.xfail_not != ''
-      run: echo "LIT_XFAIL_NOT=${{inputs.xfail_not}}" >> $GITHUB_ENV
-
     - name: Set LIT_XFAIL
       if: inputs.xfail != ''
       run: echo "LIT_XFAIL=${{inputs.xfail}}" >> $GITHUB_ENV
@@ -181,6 +177,10 @@ jobs:
       if: inputs.filter_out != ''
       run: echo "LIT_FILTER_OUT=${{inputs.filter_out}}" >> $GITHUB_ENV
 
+    - name: Set LIT_XFAIL_NOT
+      if: inputs.xfail_not != ''
+      run: echo "LIT_XFAIL_NOT=${{inputs.xfail_not}}" >> $GITHUB_ENV
+
     # TODO: remove once intel/llvm lit tests can properly recognize the GPU
     - name: Configure hardware platform feature for L0
       if: matrix.adapter.name == 'L0'

.github/workflows/e2e_cuda.yml

@@ -21,3 +21,4 @@ jobs:
       config: "--cuda"
       unit: "gpu"
       extra_lit_flags: "-sv --max-time=3600"
+      xfail: "Regression/device_num.cpp"

.github/workflows/e2e_level_zero.yml

@@ -21,9 +21,11 @@ jobs:
       config: ""
       unit: "gpu"
       # Failing tests
-      xfail: "DeviceCodeSplit/grf.cpp;ESIMD/grf.cpp;ESIMD/mask_expand_load.cpp;KernelAndProgram/target_register_alloc_mode.cpp;Matrix/SG32/get_coord_int8_matB.cpp;Matrix/get_coord_int8_matB.cpp;Matrix/joint_matrix_prefetch.cpp;Matrix/joint_matrix_rowmajorA_rowmajorB.cpp;  SYCL :: ESIMD/mask_expand_load.cpp;Matrix/SPVCooperativeMatrix/joint_matrix_bf16_fill_k_cache_OOB.cpp;Matrix/SPVCooperativeMatrix/joint_matrix_out_bounds.cpp;Matrix/SPVCooperativeMatrix/joint_matrix_prefetch.cpp;Matrix/SPVCooperativeMatrix/joint_matrix_rowmajorA_rowmajorB.cpp;Matrix/element_wise_all_ops_1d.cpp;Matrix/element_wise_all_ops_1d_cont.cpp;Matrix/element_wise_all_ops_scalar.cpp;Matrix/joint_matrix_bf16_fill_k_cache_OOB.cpp;Matrix/joint_matrix_bf16_fill_k_cache_prefetch.cpp;Matrix/joint_matrix_out_bounds.cpp;Matrix/joint_matrix_unaligned_k.cpp"
+      xfail: "InvokeSimd/Regression/call_vadd_1d_spill.cpp;InvokeSimd/Regression/ImplicitSubgroup/call_vadd_1d_spill.cpp;ESIMD/mask_expand_load.cpp;Matrix/joint_matrix_prefetch.cpp;ESIMD/mask_expand_load.cpp;Matrix/SPVCooperativeMatrix/joint_matrix_prefetch.cpp;Matrix/joint_matrix_bf16_fill_k_cache_prefetch.cpp;Matrix/SPVCooperativeMatrix/element_wise_ops.cpp;"
+      # Unexpectedly Passed Tests
+      xfail_not: ""
       # Flaky tests
-      filter_out: "ESIMD/named_barriers/loop_extended.cpp;ESIMD/local_accessor_copy_to_from.cpp;"
+      filter_out: "Basic/accessor/accessor.cpp|DeviceArchitecture/device_architecture_comparison_on_device_aot.cpp|Graph/Explicit/interop-level-zero-launch-kernel.cpp|Graph/RecordReplay/interop-level-zero-launch-kernel.cpp|syclcompat/launch/launch_policy_lmem.cpp"
       # These runners by default spawn upwards of 260 workers.
       # We also add a time out just in case some test hangs
       extra_lit_flags: "--param gpu-intel-pvc=True --param gpu-intel-pvc-1T=True -sv -j 100 --max-time=3600"

.github/workflows/e2e_opencl.yml

@@ -20,5 +20,5 @@ jobs:
       prefix: ""
       config: ""
       unit: "cpu"
-      xfail: "AOT/double.cpp;AOT/half.cpp;AOT/reqd-sg-size.cpp;Basic/built-ins/marray_geometric.cpp;KernelCompiler/kernel_compiler_spirv.cpp;KernelCompiler/opencl_queries.cpp"
+      xfail: "AOT/double.cpp;AOT/half.cpp;AOT/reqd-sg-size.cpp;Basic/built-ins/marray_geometric.cpp;KernelCompiler/kernel_compiler_spirv.cpp;KernelCompiler/opencl_queries.cpp;NonUniformGroups/ballot_group.cpp;NonUniformGroups/ballot_group_algorithms.cpp;NonUniformGroups/fixed_size_group_algorithms.cpp;NonUniformGroups/opportunistic_group.cpp;NonUniformGroups/opportunistic_group_algorithms.cpp;NonUniformGroups/tangle_group.cpp;NonUniformGroups/tangle_group_algorithms.cpp"
       extra_lit_flags: "-sv --max-time=3600"

.github/workflows/multi_device.yml

@@ -30,12 +30,11 @@ jobs:
     - name: Install pip packages
       run: pip install -r third_party/requirements.txt
 
-    # TODO: enable once test failure are fixed/ignored
-    # - name: Download DPC++
-    #   run: |
-    #     wget -O ${{github.workspace}}/dpcpp_compiler.tar.gz https://github.com/intel/llvm/releases/download/nightly-2024-01-29/sycl_linux.tar.gz
-    #     mkdir dpcpp_compiler
-    #     tar -xvf ${{github.workspace}}/dpcpp_compiler.tar.gz -C dpcpp_compiler
+    - name: Download DPC++
+      run: |
+        wget -O ${{github.workspace}}/dpcpp_compiler.tar.gz https://github.com/intel/llvm/releases/download/nightly-2024-01-29/sycl_linux.tar.gz
+        mkdir dpcpp_compiler
+        tar -xvf ${{github.workspace}}/dpcpp_compiler.tar.gz -C dpcpp_compiler
 
     - name: Configure CMake
       shell: bash -el {0}
@@ -49,6 +48,8 @@ jobs:
         -DUR_BUILD_TESTS=ON
         -DUR_BUILD_ADAPTER_${{matrix.adapter.name}}=ON
         -DUR_TEST_DEVICES_COUNT=2
+        -DUR_DPCXX=${{github.workspace}}/dpcpp_compiler/bin/clang++
+        -DUR_SYCL_LIBRARY_DIR=${{github.workspace}}/dpcpp_compiler/lib
 
     - name: Build
       run: cmake --build ${{github.workspace}}/build -j $(nproc)
@@ -60,4 +61,4 @@ jobs:
 
     - name: Test adapters
       working-directory: ${{github.workspace}}/build
-      run: env UR_CTS_ADAPTER_PLATFORM="${{matrix.adapter.platform}}" ctest -C ${{matrix.build_type}} --output-on-failure -L "conformance" --timeout 180
+      run: env UR_CTS_ADAPTER_PLATFORM="${{matrix.adapter.platform}}" ctest -C ${{matrix.build_type}} --output-on-failure -L "conformance" -E "enqueue|kernel|program|integration|exp_command_buffer|exp_enqueue_native|exp_launch_properties|exp_usm_p2p" --timeout 180

.github/workflows/source-checks.yml

@@ -0,0 +1,49 @@
+name: Source Checks
+
+on: [push, pull_request]
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+jobs:
+  source-checks:
+    name: Source Checks
+    # TODO: Make this run on Windows (#888)
+    runs-on: 'ubuntu-22.04'
+
+    steps:
+    - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
+
+    - name: Install apt packages
+      run: |
+        sudo apt-get update
+        sudo apt-get install -y doxygen
+
+    - name: Setup PATH
+      run: echo "$HOME/.local/bin" >> $GITHUB_PATH
+
+    - name: Install pip packages
+      run: pip install -r third_party/requirements.txt
+
+    # Required for CMake
+    - name: Install libhwloc
+      run: .github/scripts/install_hwloc.sh
+
+    - name: Configure CMake
+      run: >
+        cmake
+        -B${{github.workspace}}/build
+        -DUR_ENABLE_TRACING=OFF
+        -DCMAKE_BUILD_TYPE=Debug
+        -DUR_BUILD_TESTS=OFF
+        -DUR_FORMAT_CPP_STYLE=ON
+
+    - name: Verify that each source file contains a license
+      run: cmake --build ${{github.workspace}}/build --target verify-licenses
+
+    - name: Generate source from spec, check for uncommitted diff
+      run: cmake --build ${{github.workspace}}/build --target check-generated