Merge branch 'sycl' into commonhpp_reference_calculation_change

Author: ggojska

.github/workflows/release-binaries.yml

@@ -301,7 +301,7 @@ jobs:
 
     - name: Attest Build Provenance
       id: provenance
-      uses: actions/attest-build-provenance@db473fddc028af60658334401dc6fa3ffd8669fd # v2.3.0
+      uses: actions/attest-build-provenance@e8998f949152b193b063cb0ec769d69d929409be # v2.4.0
       with:
         subject-path: ${{ needs.prepare.outputs.release-binary-filename }}
 

.github/workflows/release-sources.yml

@@ -92,7 +92,7 @@ jobs:
       - name: Attest Build Provenance
         if: github.event_name != 'pull_request'
         id: provenance
-        uses: actions/attest-build-provenance@db473fddc028af60658334401dc6fa3ffd8669fd # v2.3.0
+        uses: actions/attest-build-provenance@e8998f949152b193b063cb0ec769d69d929409be # v2.4.0
         with:
           subject-path: "*.xz"
       - if: github.event_name != 'pull_request'

.github/workflows/sycl-nightly-benchmarking.yml

@@ -0,0 +1,11 @@
+name: SYCL Nightly Benchmarking
+
+# Placeholder workflow for testing https://github.com/intel/llvm/pull/19280
+on:
+  workflow_dispatch:
+
+jobs:
+  do-nothing:
+    runs-on: ubuntu-latest
+    steps:
+      - run: echo "Test workflow for 19280"

clang/include/clang/Sema/SemaSYCL.h

@@ -265,6 +265,8 @@ class SemaSYCL : public SemaBase {
 
   llvm::DenseSet<const FunctionDecl *> SYCLKernelFunctions;
 
+  llvm::DenseSet<const FunctionDecl *> FreeFunctionDeclarations;
+
 public:
   SemaSYCL(Sema &S);
 
@@ -357,7 +359,9 @@ class SemaSYCL : public SemaBase {
   void ConstructOpenCLKernel(FunctionDecl *KernelCallerFunc, MangleContext &MC);
   void SetSYCLKernelNames();
   void MarkDevices();
+  void processFreeFunctionDeclaration(const FunctionDecl *FD);
   void ProcessFreeFunction(FunctionDecl *FD);
+  void finalizeFreeFunctionKernels();
 
   /// Get the number of fields or captures within the parsed type.
   ExprResult ActOnSYCLBuiltinNumFieldsExpr(ParsedType PT);

clang/lib/Basic/Targets.cpp

@@ -119,10 +119,6 @@ std::unique_ptr<TargetInfo> AllocateTarget(const llvm::Triple &Triple,
     return nullptr;
 
   case llvm::Triple::UnknownArch:
-    // native_cpu is only known to Clang, not to LLVM.
-    if (Triple.str() == "native_cpu")
-      return std::make_unique<NativeCPUTargetInfo>(Triple, Opts);
-
     return nullptr;
 
   case llvm::Triple::arc:
@@ -843,6 +839,9 @@ std::unique_ptr<TargetInfo> AllocateTarget(const llvm::Triple &Triple,
 
   case llvm::Triple::xtensa:
     return std::make_unique<XtensaTargetInfo>(Triple, Opts);
+
+  case llvm::Triple::native_cpu:
+    return std::make_unique<NativeCPUTargetInfo>(Triple, Opts);
   }
 }
 } // namespace targets

clang/lib/CodeGen/CodeGenFunction.cpp

@@ -1932,7 +1932,7 @@ void CodeGenFunction::GenerateCode(GlobalDecl GD, llvm::Function *Fn,
     // Implicit copy-assignment gets the same special treatment as implicit
     // copy-constructors.
     emitImplicitAssignmentOperatorBody(Args);
-  } else if (FD->hasAttr<OpenCLKernelAttr>() &&
+  } else if (getLangOpts().OpenCL && FD->hasAttr<OpenCLKernelAttr>() &&
              GD.getKernelReferenceKind() == KernelReferenceKind::Kernel) {
     CallArgList CallArgs;
     for (unsigned i = 0; i < Args.size(); ++i) {

clang/lib/CodeGen/CodeGenModule.cpp

@@ -4341,7 +4341,8 @@ void CodeGenModule::EmitGlobal(GlobalDecl GD) {
 
   // Ignore declarations, they will be emitted on their first use.
   if (const auto *FD = dyn_cast<FunctionDecl>(Global)) {
-    if (FD->hasAttr<OpenCLKernelAttr>() && FD->doesThisDeclarationHaveABody())
+    if (LangOpts.OpenCL && FD->hasAttr<OpenCLKernelAttr>() &&
+        FD->doesThisDeclarationHaveABody())
       addDeferredDeclToEmit(GlobalDecl(FD, KernelReferenceKind::Stub));
 
     // Update deferred annotations with the latest declaration if the function

clang/lib/CodeGen/TargetInfo.cpp

@@ -119,10 +119,6 @@ unsigned TargetCodeGenInfo::getOpenCLKernelCallingConv() const {
 
 void TargetCodeGenInfo::setOCLKernelStubCallingConvention(
     const FunctionType *&FT) const {
-
-  if (getABIInfo().getContext().getLangOpts().SYCLIsNativeCPU)
-     return;
-
   FT = getABIInfo().getContext().adjustFunctionType(
       FT, FT->getExtInfo().withCallingConv(CC_C));
 }

clang/lib/CodeGen/Targets/AMDGPU.cpp

@@ -305,8 +305,6 @@ class AMDGPUTargetCodeGenInfo : public TargetCodeGenInfo {
   void setTargetAttributes(const Decl *D, llvm::GlobalValue *GV,
                            CodeGen::CodeGenModule &M) const override;
   unsigned getOpenCLKernelCallingConv() const override;
-  void
-  setOCLKernelStubCallingConvention(const FunctionType *&FT) const override;
 
   llvm::Constant *getNullPointer(const CodeGen::CodeGenModule &CGM,
       llvm::PointerType *T, QualType QT) const override;
@@ -437,14 +435,6 @@ unsigned AMDGPUTargetCodeGenInfo::getOpenCLKernelCallingConv() const {
   return llvm::CallingConv::AMDGPU_KERNEL;
 }
 
-void AMDGPUTargetCodeGenInfo::setOCLKernelStubCallingConvention(
-    const FunctionType *&FT) const {
-  bool IsSYCL = getABIInfo().getContext().getLangOpts().isSYCL();
-  FT = getABIInfo().getContext().adjustFunctionType(
-      FT,
-      FT->getExtInfo().withCallingConv(!IsSYCL ? CC_C : CC_AMDGPUKernelCall));
-}
-
 // Currently LLVM assumes null pointers always have value 0,
 // which results in incorrectly transformed IR. Therefore, instead of
 // emitting null pointers in private and local address spaces, a null

clang/lib/CodeGen/Targets/SPIR.cpp

@@ -394,9 +394,6 @@ void SPIRVTargetCodeGenInfo::setCUDAKernelCallingConvention(
 
 void CommonSPIRTargetCodeGenInfo::setOCLKernelStubCallingConvention(
     const FunctionType *&FT) const {
-  // Disable kernel stub for sycl
-  if (getABIInfo().getContext().getLangOpts().isSYCL())
-     return;
   FT = getABIInfo().getContext().adjustFunctionType(
       FT, FT->getExtInfo().withCallingConv(CC_SpirFunction));
 }