Merge remote-tracking branch 'origin/sycl' into private/asachkov/accessor-iterator

Author: AlexeySachkov

.github/CODEOWNERS

@@ -63,6 +63,8 @@ llvm/lib/SYCLLowerIR/CMakeLists.txt @intel/dpcpp-tools-reviewers @intel/dpcpp-es
 # invoke_simd
 sycl/include/sycl/ext/oneapi/experimental/invoke_simd.hpp @rolandschulz @kbobrovs
 sycl/include/std/experimental/simd.hpp @rolandschulz @kbobrovs
+llvm/lib/SYCLLowerIR/LowerInvokeSimd.cpp @kbobrovs @v-klochkov @rolandschulz
+llvm/include/llvm/SYCLLowerIR/LowerInvokeSimd.h @kbobrovs @v-klochkov @rolandschulz
 
 # DevOps configs
 .github/workflows/ @intel/dpcpp-devops-reviewers

.github/workflows/sycl_gen_test_matrix.yml

@@ -19,6 +19,10 @@ on:
         type: string
         required: true
         default: ""
+      cts_config:
+        type: string
+        required: false
+        default: ""
       uniq:
         description: Unique string to name dynamic runners in AWS
         type: string
@@ -33,6 +37,9 @@ on:
       lts_matrix:
         description: "Generated Matrix"
         value: ${{ jobs.test_matrix.outputs.lts_matrix }}
+      cts_matrix:
+        description: "Generated SYCL CTS Matrix"
+        value: ${{ jobs.test_matrix.outputs.cts_matrix }}
       lts_aws_matrix:
         description: "Generated Matrix AWS subset"
         value: ${{ jobs.test_matrix.outputs.lts_aws_matrix }}
@@ -42,6 +49,7 @@ jobs:
     runs-on: ubuntu-latest
     outputs:
       lts_matrix: ${{ steps.work.outputs.lts_matrix }}
+      cts_matrix: ${{ steps.work.outputs.cts_matrix }}
       lts_aws_matrix: ${{ steps.work.outputs.lts_aws_matrix }}
     steps:
     - name: Download scripts and configs

.github/workflows/sycl_linux_build_and_test.yml

@@ -56,6 +56,18 @@ on:
         type: number
         required: false
         default: 4
+      cts_matrix:
+        type: string
+        required: false
+        default: ""
+      cts_cmake_extra_args:
+        type: string
+        required: false
+        default: ""
+      cts_ref:
+        type: string
+        required: false
+        default: 'SYCL-2020'
 
 jobs:
   build:
@@ -230,6 +242,53 @@ jobs:
         check_sycl_all: ${{ matrix.check_sycl_all }}
         results_name_suffix: ${{ matrix.config }}_${{ inputs.build_artifact_suffix }}
         cmake_args: '${{ matrix.cmake_args }} ${{ inputs.lts_cmake_extra_args }}'
+  
+  khronos_sycl_cts:
+    needs: build
+    if: ${{ inputs.cts_matrix != '' }}
+    strategy:
+      fail-fast: false
+      max-parallel: ${{ inputs.max_parallel }}
+      matrix:
+        include: ${{ fromJSON(inputs.cts_matrix) }}
+    name: ${{ matrix.name }}
+    runs-on: ${{ matrix.runs-on }}
+    env: ${{ matrix.env }}
+    container:
+      image: ${{ matrix.image }}
+      options: ${{ matrix.container_options }}
+    steps:
+    - name: Reset GPU
+      if: ${{ contains(matrix.config, 'gen9') }}
+      run: |
+        sudo mount -t debugfs none /sys/kernel/debug
+        sudo bash -c 'echo 1 > /sys/kernel/debug/dri/0/i915_wedged'
+    - run: cp -r /actions .
+    - name: Register cleanup after job is finished
+      uses: ./actions/cleanup
+    - name: Install drivers
+      if: env.compute_runtime_tag != ''
+      run: |
+        if [ -e /opt/install_drivers.sh ]; then
+          # TODO pack this script into container
+          wget raw.githubusercontent.com/intel/llvm/${{ github.sha }}/devops/scripts/get_release.py
+          sudo mv get_release.py /opt/
+          sudo -E /opt/install_drivers.sh --all
+        fi
+    # FIXME cached_checkout fails here, but works everywhere else
+    # TODO: figure out if we remove this action
+    - uses: actions/checkout@v2
+      with:
+        path: llvm
+    # TODO should this action be packed into container as well?
+    - uses: ./llvm/devops/actions/khronos_cts_test
+      name: Run Khronos SYCL CTS
+      with:
+        test_ref: ${{ inputs.cts_ref }}
+        sycl_artifact: sycl_linux_${{ inputs.build_artifact_suffix }}
+        sycl_archive: llvm_sycl.tar.xz
+        sycl_device_filter: ${{ matrix.sycl_device_filter }}
+        cmake_args: '${{ matrix.cmake_args }} ${{ inputs.cts_cmake_extra_args }}'
 
   aws-stop:
     name: Stop AWS

.github/workflows/sycl_macos_build_and_test.yml

@@ -0,0 +1,73 @@
+name: Reusable SYCL macOS build and test workflow
+
+on:
+  workflow_call:
+    inputs:
+      build_ref:
+        type: string
+        required: false
+      build_cache_suffix:
+        type: string
+        required: false
+        default: "default"
+      build_cache_size:
+        type: string
+        required: false
+        default: 2G
+      build_configure_extra_args:
+        type: string
+        required: false
+        default: ""
+      build_artifact_suffix:
+        type: string
+        required: false
+        default: "default"
+
+jobs:
+  build:
+    name: Build
+    runs-on: macos-12
+    steps:
+    - name: Install dependencies
+      run: brew install ccache ninja
+    - uses: actions/checkout@v3
+      with:
+        ref: ${{ inputs.build_ref }}
+        path: src
+    - uses: actions/cache@v3
+      with:
+        path: build_cache_${{ inputs.build_cache_suffix }}
+        key: sycl-${{ runner.os }}-${{ inputs.build_cache_suffix }}-${{ github.sha }}
+        restore-keys: sycl-${{ runner.os }}-${{ inputs.build_cache_suffix }}-
+    - name: Configure
+      env:
+        CACHE_SUFFIX: ${{ inputs.build_cache_suffix }}
+        CACHE_SIZE: ${{ inputs.build_cache_size }}
+        ARGS: ${{ inputs.build_configure_extra_args }}
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/build_cache_$CACHE_SUFFIX
+        mkdir -p $GITHUB_WORKSPACE/build
+        cd $GITHUB_WORKSPACE/build
+        python3 $GITHUB_WORKSPACE/src/buildbot/configure.py -w $GITHUB_WORKSPACE \
+          -s $GITHUB_WORKSPACE/src -o $GITHUB_WORKSPACE/build -t Release \
+          --ci-defaults $ARGS \
+          --cmake-opt="-DLLVM_CCACHE_BUILD=ON" \
+          --cmake-opt="-DLLVM_CCACHE_DIR=$GITHUB_WORKSPACE/build_cache_$CACHE_SUFFIX" \
+          --cmake-opt="-DLLVM_CCACHE_MAXSIZE=$CACHE_SIZE" \
+          --cmake-opt="-DLLVM_INSTALL_UTILS=ON" \
+          --cmake-opt="-DSYCL_PI_TESTS=OFF"
+    - name: Compile
+      id: build
+      run: cmake --build $GITHUB_WORKSPACE/build --target sycl-toolchain
+    - name: Install
+      run: |
+        cmake --build $GITHUB_WORKSPACE/build --target deploy-sycl-toolchain
+
+    - name: Pack toolchain
+      run: tar -cJf llvm_sycl.tar.xz -C $GITHUB_WORKSPACE/build/install .
+    - name: Upload toolchain
+      uses: actions/upload-artifact@v2
+      with:
+        name: sycl_macos_${{ inputs.build_artifact_suffix }}
+        path: llvm_sycl.tar.xz
+

.github/workflows/sycl_post_commit.yml

@@ -17,6 +17,7 @@ jobs:
     uses: ./.github/workflows/sycl_gen_test_matrix.yml
     with:
       lts_config: "l0_gen9"
+      cts_config: "cuda"
   linux_default:
     name: Linux Default
     needs: test_matrix
@@ -26,6 +27,7 @@ jobs:
       build_cache_root: "/__w/llvm"
       build_artifact_suffix: "post_commit"
       lts_matrix: ${{ needs.test_matrix.outputs.lts_matrix }}
+      cts_matrix: ${{ needs.test_matrix.outputs.cts_matrix }}
       lts_aws_matrix: ${{ needs.test_matrix.outputs.lts_aws_matrix }}
   linux_no_assert:
     name: Linux (no assert)
@@ -111,3 +113,8 @@ jobs:
     name: Windows
     if: github.repository == 'intel/llvm'
     uses: ./.github/workflows/sycl_windows_build_and_test.yml
+
+  macos_default:
+    name: macOS
+    if: github.repository == 'intel/llvm'
+    uses: ./.github/workflows/sycl_macos_build_and_test.yml

buildbot/configure.py

@@ -13,10 +13,15 @@ def do_configure(args):
     if not os.path.isdir(abs_obj_dir):
       os.makedirs(abs_obj_dir)
 
-    llvm_external_projects = 'sycl;llvm-spirv;opencl;libdevice;xpti;xptifw'
+    llvm_external_projects = 'sycl;llvm-spirv;opencl;xpti;xptifw'
+
+    # libdevice build requires a working SYCL toolchain, which is not the case
+    # with macOS target right now.
+    if sys.platform != "darwin":
+        llvm_external_projects += ';libdevice'
 
     libclc_amd_target_names = ';amdgcn--;amdgcn--amdhsa'
-    libclc_nvidia_target_names = 'nvptx64--;nvptx64--nvidiacl'
+    libclc_nvidia_target_names = ';nvptx64--;nvptx64--nvidiacl'
 
     if args.llvm_external_projects:
         llvm_external_projects += ";" + args.llvm_external_projects.replace(",", ";")
@@ -39,11 +44,14 @@ def do_configure(args):
     llvm_enable_sphinx = 'OFF'
     llvm_build_shared_libs = 'OFF'
     llvm_enable_lld = 'OFF'
-    sycl_enabled_plugins = ["opencl", "level_zero"]
+    sycl_enabled_plugins = ["opencl"]
 
     sycl_enable_xpti_tracing = 'ON'
     xpti_enable_werror = 'OFF'
 
+    if sys.platform != "darwin":
+        sycl_enabled_plugins.append("level_zero")
+
     # lld is needed on Windows or for the HIP plugin on AMD
     if platform.system() == 'Windows' or (args.hip and args.hip_platform == 'AMD'):
         llvm_enable_projects += ';lld'
@@ -104,17 +112,19 @@ def do_configure(args):
 
         # For clang-format, clang-tidy and code coverage
         llvm_enable_projects += ";clang-tools-extra;compiler-rt"
-        # libclc is required for CI validation
-        if 'libclc' not in llvm_enable_projects:
-            llvm_enable_projects += ';libclc'
-        # libclc passes `--nvvm-reflect-enable=false`, build NVPTX to enable it
-        if 'NVPTX' not in llvm_targets_to_build:
-            llvm_targets_to_build += ';NVPTX'
-        # Add both NVIDIA and AMD libclc targets
-        if libclc_amd_target_names not in libclc_targets_to_build:
-            libclc_targets_to_build += libclc_amd_target_names
-        if libclc_nvidia_target_names not in libclc_targets_to_build:
-            libclc_targets_to_build += libclc_nvidia_target_names
+        if sys.platform != "darwin":
+            # libclc is required for CI validation
+            if 'libclc' not in llvm_enable_projects:
+                llvm_enable_projects += ';libclc'
+            # libclc passes `--nvvm-reflect-enable=false`, build NVPTX to enable it
+            if 'NVPTX' not in llvm_targets_to_build:
+                llvm_targets_to_build += ';NVPTX'
+            # Add both NVIDIA and AMD libclc targets
+            if libclc_amd_target_names not in libclc_targets_to_build:
+                libclc_targets_to_build += libclc_amd_target_names
+            if libclc_nvidia_target_names not in libclc_targets_to_build:
+                libclc_targets_to_build += libclc_nvidia_target_names
+            libclc_gen_remangled_variants = 'ON'
 
     if args.enable_plugin:
         sycl_enabled_plugins += args.enable_plugin

clang/include/clang/Basic/Attr.td

@@ -1429,6 +1429,14 @@ def SYCLRequiresDecomposition : InheritableAttr {
   let Documentation = [InternalOnly];
 }
 
+def SYCLGenerateNewType : InheritableAttr {
+  // No spellings, as this is for internal use.
+  let Spellings = [];
+  let Subjects = SubjectList<[Named]>;
+  let LangOpts = [SYCLIsDevice, SYCLIsHost];
+  let Documentation = [InternalOnly];
+}
+
 def SYCLIntelKernelArgsRestrict : InheritableAttr {
   let Spellings = [CXX11<"intel", "kernel_args_restrict">];
   let Subjects = SubjectList<[Function], ErrorDiag>;

clang/include/clang/Basic/CodeGenOptions.def

@@ -488,6 +488,10 @@ CODEGENOPT(SkipRaxSetup, 1, 0)
 ENUM_CODEGENOPT(ZeroCallUsedRegs, llvm::ZeroCallUsedRegs::ZeroCallUsedRegsKind,
                 5, llvm::ZeroCallUsedRegs::ZeroCallUsedRegsKind::Skip)
 
+/// Whether to expect -main-file-name to be an absolute path to use it for
+/// checksum calculations or not.
+CODEGENOPT(SYCLUseMainFileName, 1, 0)
+
 /// Whether to use opaque pointers.
 CODEGENOPT(OpaquePointers, 1, 0)
 

clang/include/clang/Basic/CodeGenOptions.h

@@ -220,6 +220,9 @@ class CodeGenOptions : public CodeGenOptionsBase {
   /// file, for example with -save-temps.
   std::string MainFileName;
 
+  /// The user provided name for the "main file", with its full path.
+  std::string FullMainFileName;
+
   /// The name for the split debug info file used for the DW_AT_[GNU_]dwo_name
   /// attribute in the skeleton CU.
   std::string SplitDwarfFile;

clang/include/clang/Driver/Options.td

@@ -6156,6 +6156,10 @@ def main_file_name : Separate<["-"], "main-file-name">,
   HelpText<"Main file name to use for debug info and source if missing">,
   Flags<[CC1Option, CC1AsOption, NoDriverOption]>,
   MarshallingInfoString<CodeGenOpts<"MainFileName">>;
+def full_main_file_name : Separate<["-"], "full-main-file-name">,
+  HelpText<"File name with full path to use for debug info during host and device compile">,
+  Flags<[CC1Option, CC1AsOption, NoDriverOption]>,
+  MarshallingInfoString<CodeGenOpts<"FullMainFileName">>;
 def split_dwarf_output : Separate<["-"], "split-dwarf-output">,
   HelpText<"File name to use for split dwarf debug info output">,
   Flags<[CC1Option, CC1AsOption, NoDriverOption]>,
@@ -6517,6 +6521,10 @@ def fsycl_disable_range_rounding : Flag<["-"], "fsycl-disable-range-rounding">,
 def fsycl_enable_int_header_diags: Flag<["-"], "fsycl-enable-int-header-diags">,
   HelpText<"Enable diagnostics that require the SYCL integration header.">,
   MarshallingInfoFlag<LangOpts<"SYCLEnableIntHeaderDiags">>;
+def fsycl_use_main_file_name : Flag<["-"], "fsycl-use-main-file-name">,
+  HelpText<"Tells compiler that -main-file-name contains an absolute path and "
+           "file specified there should be used for checksum calculation.">,
+  MarshallingInfoFlag<CodeGenOpts<"SYCLUseMainFileName">>;
 
 } // let Flags = [CC1Option, NoDriverOption]