Merge branch 'sycl' into Testing-adding-handles-to-opencl

.github/CODEOWNERS

@@ -36,6 +36,7 @@ sycl/doc/extensions/ @intel/dpcpp-specification-reviewers
 
 # Unified Runtime
 sycl/cmake/modules/FetchUnifiedRuntime.cmake @intel/unified-runtime-reviewers
+sycl/cmake/modules/UnifiedRuntimeTag.cmake @intel/unified-runtime-reviewers
 sycl/include/sycl/detail/ur.hpp @intel/unified-runtime-reviewers
 sycl/source/detail/posix_ur.cpp @intel/unified-runtime-reviewers
 sycl/source/detail/ur.cpp @intel/unified-runtime-reviewers

.github/workflows/pr-code-format.yml

@@ -7,6 +7,9 @@ on:
   pull_request:
     branches:
       - main
+      - sycl
+      - sycl-devops-pr/**
+      - sycl-rel-**
       - 'users/**'
 
 jobs:

.github/workflows/scorecard.yml

@@ -57,6 +57,6 @@ jobs:
 
       # Upload the results to GitHub's code scanning dashboard.
       - name: "Upload to code-scanning"
-        uses: github/codeql-action/upload-sarif@e2b3eafc8d227b0241d48be5f425d47c2d750a13 # v3.26.10
+        uses: github/codeql-action/upload-sarif@662472033e021d55d94146f66f6058822b0b39fd # v3.27.0
         with:
           sarif_file: results.sarif

.github/workflows/sycl-containers.yaml

@@ -51,6 +51,13 @@ jobs:
             file: ubuntu2204_intel_drivers
             tag: unstable
             build_args: "use_latest=true"
+          - name: Build + Intel Drivers Ubuntu 22.04 Docker image
+            file: ubuntu2204_intel_drivers
+            tag: alldeps
+            build_args: |
+              base_image=ghcr.io/intel/llvm/ubuntu2204_build
+              base_tag=latest
+              use_latest=false
     steps:
       - name: Checkout
         uses: actions/checkout@v4

.github/workflows/sycl-linux-build.yml

@@ -212,6 +212,12 @@ jobs:
       if: always() && !cancelled() && contains(inputs.changes, 'libdevice')
       run: |
         cmake --build $GITHUB_WORKSPACE/build --target check-libdevice
+    - name: Check E2E test requirements
+      if: always() && !cancelled() && !contains(inputs.changes, 'sycl')
+      run: |
+        # TODO consider moving this to Dockerfile.
+        export LD_LIBRARY_PATH=/usr/local/cuda/compat/:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
+        LIT_OPTS="--allow-empty-runs" LIT_FILTER="e2e_test_requirements" cmake --build $GITHUB_WORKSPACE/build --target check-sycl
     - name: Install
       if: ${{ always() && !cancelled() && steps.build.conclusion == 'success' }}
       # TODO replace utility installation with a single CMake target

.github/workflows/sycl-linux-run-tests.yml

@@ -185,8 +185,8 @@ jobs:
       with:
         path: khronos_sycl_cts
         repository: 'KhronosGroup/SYCL-CTS'
-        ref: 'SYCL-2020'
-        default_branch: 'SYCL-2020'
+        ref: 'main'
+        default_branch: 'main'
         cache_path: "/__w/repo_cache/"
     - name: SYCL CTS GIT submodules init
       if: inputs.tests_selector == 'cts'

.github/workflows/sycl-macos-build-and-test.yml

@@ -24,7 +24,7 @@ permissions: read-all
 jobs:
   build:
     name: Build
-    runs-on: macos-12
+    runs-on: macos-13
     env:
       CCACHE_DIR: $GITHUB_WORKSPACE/build_cache_${{ inputs.build_cache_suffix }}
       CCACHE_MAXSIZE: ${{ inputs.build_cache_size }}

.github/workflows/sycl-nightly.yml

@@ -198,7 +198,7 @@ jobs:
           echo "TAG=$(date +'%Y-%m-%d')-${GITHUB_SHA::7}" >> "$GITHUB_OUTPUT"
         fi
     - name: Upload binaries
-      uses: softprops/action-gh-release@c062e08bd532815e2082a85e87e3ef29c3e6d191
+      uses: softprops/action-gh-release@e7a8f85e1c67a31e6ed99a94b41bd0b71bbee6b8
       with:
         files: |
           sycl_linux.tar.gz

buildbot/configure.py

@@ -64,6 +64,7 @@ def do_configure(args):
 
     sycl_enable_xpti_tracing = "ON"
     xpti_enable_werror = "OFF"
+    llvm_enable_zstd = "ON"
 
     if sys.platform != "darwin":
         sycl_enabled_backends.append("level_zero")
@@ -177,6 +178,8 @@ def do_configure(args):
         "-DLLVM_ENABLE_PROJECTS={}".format(llvm_enable_projects),
         "-DSYCL_BUILD_PI_HIP_PLATFORM={}".format(sycl_build_pi_hip_platform),
         "-DLLVM_BUILD_TOOLS=ON",
+        "-DLLVM_ENABLE_ZSTD={}".format(llvm_enable_zstd),
+        "-DLLVM_USE_STATIC_ZSTD=ON",
         "-DSYCL_ENABLE_WERROR={}".format(sycl_werror),
         "-DCMAKE_INSTALL_PREFIX={}".format(install_dir),
         "-DSYCL_INCLUDE_TESTS=ON",  # Explicitly include all kinds of SYCL tests.

clang/include/clang/Basic/Attr.td

@@ -1566,12 +1566,12 @@ def SYCLType: InheritableAttr {
   let Subjects = SubjectList<[CXXRecord, Enum], ErrorDiag>;
   let LangOpts = [SYCLIsDevice, SilentlyIgnoreSYCLIsHost];
   let Args = [EnumArgument<"Type", "SYCLType", /*is_string=*/true,
-                           ["accessor", "local_accessor",
+                           ["accessor", "local_accessor", "work_group_memory",
                             "specialization_id", "kernel_handler", "buffer_location",
                             "no_alias", "accessor_property_list", "group",
                             "private_memory", "aspect", "annotated_ptr", "annotated_arg",
                             "stream", "sampler", "host_pipe", "multi_ptr"],
-                           ["accessor", "local_accessor",
+                           ["accessor", "local_accessor", "work_group_memory",
                             "specialization_id", "kernel_handler", "buffer_location",
                             "no_alias", "accessor_property_list", "group",
                             "private_memory", "aspect", "annotated_ptr", "annotated_arg",

clang/include/clang/Basic/DiagnosticDriverKinds.td

@@ -398,6 +398,10 @@ def err_drv_sycl_missing_amdgpu_arch : Error<
   "missing AMDGPU architecture for SYCL offloading; specify it with '-Xsycl-target-backend%select{|=%1}0 --offload-arch=<arch-name>'">;
 def err_drv_sycl_thinlto_split_off: Error<
   "'%0' is not supported when '%1' is set with '-fsycl'">;
+def err_drv_sycl_offload_arch_new_driver: Error<
+  "'--offload-arch' is supported when '-fsycl' is set with '--offload-new-driver'">;
+def err_drv_sycl_offload_arch_missing_value : Error<
+  "must pass in an explicit cpu or gpu architecture to '--offload-arch'">;
 def warn_drv_sycl_offload_target_duplicate : Warning<
   "SYCL offloading target '%0' is similar to target '%1' already specified; "
   "will be ignored">, InGroup<SyclTarget>;

clang/include/clang/Driver/Action.h

@@ -698,6 +698,14 @@ class OffloadWrapperJobAction : public JobAction {
   // Get the compilation step setting.
   bool getCompileStep() const { return CompileStep; }
 
+  // Set the individual wrapping setting.  This is used to tell the wrapper job
+  // action that the wrapping (and subsequent compile step) should be done
+  // with for-each instead of using -batch.
+  void setWrapIndividualFiles() { WrapIndividualFiles = true; }
+
+  // Get the individual wrapping setting.
+  bool getWrapIndividualFiles() const { return WrapIndividualFiles; }
+
   // Set the offload kind for the current wrapping job action.  Default usage
   // is to use the kind of the current toolchain.
   void setOffloadKind(OffloadKind SetKind) { Kind = SetKind; }
@@ -707,6 +715,7 @@ class OffloadWrapperJobAction : public JobAction {
 
 private:
   bool CompileStep = true;
+  bool WrapIndividualFiles = false;
   OffloadKind Kind = OFK_None;
 };
 

clang/include/clang/Driver/Driver.h

@@ -586,9 +586,9 @@ class Driver {
   /// @name Helper Methods
   /// @{
 
-  /// MakeSYCLDeviceTriple - Returns the SYCL device triple for the
+  /// getSYCLDeviceTriple - Returns the SYCL device triple for the
   /// specified subarch
-  llvm::Triple MakeSYCLDeviceTriple(StringRef TargetArch = "spir64") const;
+  llvm::Triple getSYCLDeviceTriple(StringRef TargetArch = "spir64") const;
 
   /// PrintActions - Print the list of actions.
   void PrintActions(const Compilation &C) const;

clang/include/clang/Driver/Options.td

@@ -4658,10 +4658,14 @@ def image__base : Separate<["-"], "image_base">;
 def include_ : JoinedOrSeparate<["-", "--"], "include">, Group<clang_i_Group>, EnumName<"include">,
     MetaVarName<"<file>">, HelpText<"Include file before parsing">,
     Visibility<[ClangOption, CC1Option]>;
-def include_footer : Separate<["-"], "include-footer">, Group<clang_i_Group>,
+def include_internal_footer : Separate<["-"], "include-internal-footer">, Group<clang_i_Group>,
     Visibility<[CC1Option]>,
     HelpText<"Name of the footer integration file">,  MetaVarName<"<file>">,
     MarshallingInfoString<PreprocessorOpts<"IncludeFooter">>;
+def include_internal_header : Separate<["-"], "include-internal-header">, Group<clang_i_Group>,
+    Visibility<[CC1Option]>,
+    HelpText<"Name of the header integration file">,  MetaVarName<"<file>">,
+    MarshallingInfoString<PreprocessorOpts<"IncludeHeader">>;
 def include_pch : Separate<["-"], "include-pch">, Group<clang_i_Group>,
   Visibility<[ClangOption, CC1Option]>,
   HelpText<"Include precompiled header file">, MetaVarName<"<file>">,

clang/include/clang/Driver/ToolChain.h

@@ -781,7 +781,7 @@ class ToolChain {
   virtual void AddHIPIncludeArgs(const llvm::opt::ArgList &DriverArgs,
                                  llvm::opt::ArgStringList &CC1Args) const;
 
-  /// Add arguments to use SYCL specific includes.
+  /// Add arguments to use system-specific SYCL includes.
   virtual void AddSYCLIncludeArgs(const llvm::opt::ArgList &DriverArgs,
                                   llvm::opt::ArgStringList &CC1Args) const;
 

clang/include/clang/Lex/PreprocessorOptions.h

@@ -68,6 +68,7 @@ class PreprocessorOptions {
   std::vector<std::pair<std::string, bool/*isUndef*/>> Macros;
   std::vector<std::string> Includes;
   std::string IncludeFooter;
+  std::string IncludeHeader;
   std::vector<std::string> MacroIncludes;
 
   /// Perform extra checks when loading PCM files for mutable file systems.

clang/include/clang/Sema/SemaSYCL.h

@@ -62,7 +62,8 @@ class SYCLIntegrationHeader {
     kind_pointer,
     kind_specialization_constants_buffer,
     kind_stream,
-    kind_last = kind_stream
+    kind_work_group_memory,
+    kind_last = kind_work_group_memory
   };
 
 public:

clang/lib/Driver/Compilation.cpp

@@ -344,7 +344,7 @@ static bool ActionFailed(const Action *A,
 
   // CUDA/HIP/SYCL can have the same input source code compiled multiple times
   // so do not compile again if there are already failures. It is OK to abort
-  // the CUDA pipeline on errors.
+  // the CUDA/HIP/SYCL pipeline on errors.
   if (A->isOffloading(Action::OFK_Cuda) || A->isOffloading(Action::OFK_HIP) ||
       A->isOffloading(Action::OFK_SYCL))
     return true;