Merge branch 'sycl' into iault/unsampled_image_arrays

.github/workflows/scorecard.yml

@@ -2,7 +2,7 @@
 # by a third-party and are governed by separate terms of service, privacy
 # policy, and support documentation.
 
-# Check current LLVM-Project results here: https://securityscorecards.dev/viewer/?uri=github.com/llvm/llvm-project
+# Check current LLVM-Project results here: https://securityscorecards.dev/viewer/?uri=github.com/intel/llvm
 
 name: Scorecard supply-chain security
 on:
@@ -12,7 +12,7 @@ on:
   # To guarantee Maintained check is occasionally updated. See
   # https://github.com/ossf/scorecard/blob/main/docs/checks.md#maintained
   schedule:
-    - cron: '38 20 * * *'
+    - cron: '30 20 * * *'
 
 # Declare default permissions as read only.
 permissions:
@@ -22,7 +22,7 @@ jobs:
   analysis:
     name: Scorecard analysis
     runs-on: ubuntu-latest
-    if: github.repository == 'llvm/llvm-project'
+    if: github.repository == 'intel/llvm'
     permissions:
       # Needed to upload the results to code-scanning dashboard.
       security-events: write

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

@@ -109,21 +109,37 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - name: Perf tests on Intel GEN12 Graphics system
+          - name: Intel GEN12 Graphics system
             runner: '["Linux", "gen12"]'
-          - name: Perf tests on Intel Arc A-Series Graphics system
+            image: ghcr.io/intel/llvm/ubuntu2204_intel_drivers:latest
+            image_extra_opts: --device=/dev/dri
+            reset_gpu: true
+          - name: Intel Arc A-Series Graphics system
             runner: '["Linux", "arc"]'
+            image: ghcr.io/intel/llvm/ubuntu2204_intel_drivers:latest
+            image_extra_opts: --device=/dev/dri
+            reset_gpu: true
+          - name: AMD system
+            runner: '["Linux", "amdgpu"]'
+            image: ghcr.io/intel/llvm/ubuntu2204_build:latest
+            image_extra_opts: --device=/dev/dri --device=/dev/kfd
+            extra_cmake_args: -DHIP_PLATFORM="AMD" -DAMD_ARCH="gfx1031"
+          - name: CUDA system
+            runner: '["Linux", "cuda"]'
+            image: ghcr.io/intel/llvm/ubuntu2204_build:latest
+            image_extra_opts: --gpus all
     uses: ./.github/workflows/sycl-linux-run-tests.yml
     with:
-      name: ${{ matrix.name }}
+      name: Perf tests on ${{ matrix.name }}
       runner: ${{ matrix. runner }}
-      image: ghcr.io/intel/llvm/ubuntu2204_intel_drivers:latest
-      image_options: -u 1001 --device=/dev/dri --privileged --cap-add SYS_ADMIN
+      image: ${{ matrix.image }}
+      image_options: -u 1001 --privileged --cap-add SYS_ADMIN ${{ matrix.image_extra_opts }}
       target_devices: all
-      reset_gpu: true
+      reset_gpu: ${{ matrix.reset_gpu }}
 
       env: '{"LIT_FILTER":"PerformanceTests/"}'
       extra_lit_opts: -a -j 1 --param enable-perf-tests=True
+      extra_cmake_args: ${{ matrix.extra_cmake_args }}
 
       ref: ${{ github.sha }}
       merge_ref: ''

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

@@ -100,7 +100,7 @@ on:
         options:
           - 'opencl:cpu'
           - 'opencl:gpu'
-          - 'opencl:acc'
+          - 'opencl:fpga'
           - 'ext_oneapi_level_zero:gpu'
           - 'ext_oneapi_hip:gpu'
       tests_selector:

.github/workflows/sycl-post-commit.yml

@@ -45,6 +45,7 @@ jobs:
           - name: Intel GEN12 Graphics with Level Zero
             runner: '["Linux", "gen12"]'
             extra_lit_opts: --param gpu-intel-gen12=True
+            target_devices: ext_oneapi_level_zero:gpu;opencl:fpga
           - name: Intel Arc A-Series Graphics with Level Zero
             runner: '["Linux", "arc"]'
             extra_lit_opts: --param matrix-xmx8=True --param gpu-intel-dg2=True

clang/lib/CodeGen/CGVTT.cpp

@@ -42,8 +42,8 @@ CodeGenVTables::EmitVTTDefinition(llvm::GlobalVariable *VTT,
                                   llvm::GlobalVariable::LinkageTypes Linkage,
                                   const CXXRecordDecl *RD) {
   VTTBuilder Builder(CGM.getContext(), RD, /*GenerateDefinition=*/true);
-  llvm::ArrayType *ArrayType =
-      llvm::ArrayType::get(CGM.DefaultInt8PtrTy, Builder.getVTTComponents().size());
+  llvm::ArrayType *ArrayType = llvm::ArrayType::get(
+      CGM.GlobalsInt8PtrTy, Builder.getVTTComponents().size());
 
   SmallVector<llvm::GlobalVariable *, 8> VTables;
   SmallVector<VTableAddressPointsMapTy, 8> VTableAddressPoints;
@@ -81,9 +81,6 @@ CodeGenVTables::EmitVTTDefinition(llvm::GlobalVariable *VTT,
          VTable->getValueType(), VTable, Idxs, /*InBounds=*/true,
          /*InRangeIndex=*/1);
 
-     Init = llvm::ConstantExpr::getPointerBitCastOrAddrSpaceCast(
-         Init, CGM.Int8PtrTy);
-
      VTTComponents.push_back(Init);
   }
 
@@ -117,9 +114,9 @@ llvm::GlobalVariable *CodeGenVTables::GetAddrOfVTT(const CXXRecordDecl *RD) {
 
   VTTBuilder Builder(CGM.getContext(), RD, /*GenerateDefinition=*/false);
 
-  llvm::ArrayType *ArrayType =
-    llvm::ArrayType::get(CGM.Int8PtrTy, Builder.getVTTComponents().size());
-  llvm::Align Align = CGM.getDataLayout().getABITypeAlign(CGM.Int8PtrTy);
+  llvm::ArrayType *ArrayType = llvm::ArrayType::get(
+      CGM.GlobalsInt8PtrTy, Builder.getVTTComponents().size());
+  llvm::Align Align = CGM.getDataLayout().getABITypeAlign(CGM.GlobalsInt8PtrTy);
 
   llvm::GlobalVariable *GV = CGM.CreateOrReplaceCXXRuntimeVariable(
       Name, ArrayType, llvm::GlobalValue::ExternalLinkage, Align);

clang/lib/CodeGen/CGVTables.cpp

@@ -692,7 +692,7 @@ bool CodeGenVTables::useRelativeLayout() const {
 llvm::Type *CodeGenModule::getVTableComponentType() const {
   if (UseRelativeLayout(*this))
     return Int32Ty;
-  return Int8PtrTy;
+  return GlobalsInt8PtrTy;
 }
 
 llvm::Type *CodeGenVTables::getVTableComponentType() const {
@@ -704,7 +704,7 @@ static void AddPointerLayoutOffset(const CodeGenModule &CGM,
                                    CharUnits offset) {
   builder.add(llvm::ConstantExpr::getIntToPtr(
       llvm::ConstantInt::get(CGM.PtrDiffTy, offset.getQuantity()),
-      CGM.Int8PtrTy));
+      CGM.GlobalsInt8PtrTy));
 }
 
 static void AddRelativeLayoutOffset(const CodeGenModule &CGM,
@@ -741,7 +741,7 @@ void CodeGenVTables::addVTableComponent(ConstantArrayBuilder &builder,
                                   vtableHasLocalLinkage,
                                   /*isCompleteDtor=*/false);
     else
-      return builder.add(llvm::ConstantExpr::getBitCast(rtti, CGM.Int8PtrTy));
+      return builder.add(rtti);
 
   case VTableComponent::CK_FunctionPointer:
   case VTableComponent::CK_CompleteDtorPointer:
@@ -760,7 +760,8 @@ void CodeGenVTables::addVTableComponent(ConstantArrayBuilder &builder,
               ? MD->hasAttr<CUDADeviceAttr>()
               : (MD->hasAttr<CUDAHostAttr>() || !MD->hasAttr<CUDADeviceAttr>());
       if (!CanEmitMethod)
-        return builder.add(llvm::ConstantExpr::getNullValue(CGM.Int8PtrTy));
+        return builder.add(
+            llvm::ConstantExpr::getNullValue(CGM.GlobalsInt8PtrTy));
       // Method is acceptable, continue processing as usual.
     }
 
@@ -773,20 +774,20 @@ void CodeGenVTables::addVTableComponent(ConstantArrayBuilder &builder,
       // with the local symbol. As a temporary solution, fill these components
       // with zero. We shouldn't be calling these in the first place anyway.
       if (useRelativeLayout())
-        return llvm::ConstantPointerNull::get(CGM.Int8PtrTy);
+        return llvm::ConstantPointerNull::get(CGM.GlobalsInt8PtrTy);
 
       // For NVPTX devices in OpenMP emit special functon as null pointers,
       // otherwise linking ends up with unresolved references.
       if (CGM.getLangOpts().OpenMP && CGM.getLangOpts().OpenMPIsTargetDevice &&
           CGM.getTriple().isNVPTX())
-        return llvm::ConstantPointerNull::get(CGM.Int8PtrTy);
+        return llvm::ConstantPointerNull::get(CGM.GlobalsInt8PtrTy);
       llvm::FunctionType *fnTy =
           llvm::FunctionType::get(CGM.VoidTy, /*isVarArg=*/false);
       llvm::Constant *fn = cast<llvm::Constant>(
           CGM.CreateRuntimeFunction(fnTy, name).getCallee());
       if (auto f = dyn_cast<llvm::Function>(fn))
         f->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
-      return llvm::ConstantExpr::getBitCast(fn, CGM.Int8PtrTy);
+      return fn;
     };
 
     llvm::Constant *fnPtr;
@@ -824,15 +825,26 @@ void CodeGenVTables::addVTableComponent(ConstantArrayBuilder &builder,
       return addRelativeComponent(
           builder, fnPtr, vtableAddressPoint, vtableHasLocalLinkage,
           component.getKind() == VTableComponent::CK_CompleteDtorPointer);
-    } else
-      return builder.add(llvm::ConstantExpr::getBitCast(fnPtr, CGM.Int8PtrTy));
+    } else {
+      // TODO: this icky and only exists due to functions being in the generic
+      //       address space, rather than the global one, even though they are
+      //       globals;  fixing said issue might be intrusive, and will be done
+      //       later.
+      unsigned FnAS = fnPtr->getType()->getPointerAddressSpace();
+      unsigned GVAS = CGM.GlobalsInt8PtrTy->getPointerAddressSpace();
+
+      if (FnAS != GVAS)
+        fnPtr =
+            llvm::ConstantExpr::getAddrSpaceCast(fnPtr, CGM.GlobalsInt8PtrTy);
+      return builder.add(fnPtr);
+    }
   }
 
   case VTableComponent::CK_UnusedFunctionPointer:
     if (useRelativeLayout())
       return builder.add(llvm::ConstantExpr::getNullValue(CGM.Int32Ty));
     else
-      return builder.addNullPointer(CGM.Int8PtrTy);
+      return builder.addNullPointer(CGM.GlobalsInt8PtrTy);
   }
 
   llvm_unreachable("Unexpected vtable component kind");

clang/lib/CodeGen/CodeGenModule.cpp

@@ -7982,7 +7982,7 @@ llvm::Constant *CodeGenModule::GetAddrOfRTTIDescriptor(QualType Ty,
   // FIXME: should we even be calling this method if RTTI is disabled
   // and it's not for EH?
   if (!shouldEmitRTTI(ForEH))
-    return llvm::Constant::getNullValue(Int8PtrTy);
+    return llvm::Constant::getNullValue(GlobalsInt8PtrTy);
 
   if (ForEH && Ty->isObjCObjectPointerType() &&
       LangOpts.ObjCRuntime.isGNUFamily())

clang/lib/CodeGen/ItaniumCXXABI.cpp

@@ -670,7 +670,7 @@ CGCallee ItaniumCXXABI::EmitLoadOfMemberFunctionPointer(
   CGF.EmitBlock(FnVirtual);
 
   // Cast the adjusted this to a pointer to vtable pointer and load.
-  llvm::Type *VTableTy = Builder.getInt8PtrTy();
+  llvm::Type *VTableTy = CGF.CGM.GlobalsInt8PtrTy;
   CharUnits VTablePtrAlign =
     CGF.CGM.getDynamicOffsetAlignment(ThisAddr.getAlignment(), RD,
                                       CGF.getPointerAlign());
@@ -1942,11 +1942,11 @@ llvm::Value *ItaniumCXXABI::getVTableAddressPointInStructorWithVTT(
   /// Load the VTT.
   llvm::Value *VTT = CGF.LoadCXXVTT();
   if (VirtualPointerIndex)
-    VTT = CGF.Builder.CreateConstInBoundsGEP1_64(
-        CGF.VoidPtrTy, VTT, VirtualPointerIndex);
+    VTT = CGF.Builder.CreateConstInBoundsGEP1_64(CGF.GlobalsVoidPtrTy, VTT,
+                                                 VirtualPointerIndex);
 
   // And load the address point from the VTT.
-  return CGF.Builder.CreateAlignedLoad(CGF.VoidPtrTy, VTT,
+  return CGF.Builder.CreateAlignedLoad(CGF.GlobalsVoidPtrTy, VTT,
                                        CGF.getPointerAlign());
 }
 
@@ -1974,12 +1974,13 @@ llvm::GlobalVariable *ItaniumCXXABI::getAddrOfVTable(const CXXRecordDecl *RD,
       CGM.getItaniumVTableContext().getVTableLayout(RD);
   llvm::Type *VTableType = CGM.getVTables().getVTableType(VTLayout);
 
-  // Use pointer alignment for the vtable. Otherwise we would align them based
-  // on the size of the initializer which doesn't make sense as only single
-  // values are read.
+  // Use pointer to global alignment for the vtable. Otherwise we would align
+  // them based on the size of the initializer which doesn't make sense as only
+  // single values are read.
+  LangAS AS = CGM.GetGlobalVarAddressSpace(nullptr);
   unsigned PAlign = CGM.getItaniumVTableContext().isRelativeLayout()
                         ? 32
-                        : CGM.getTarget().getPointerAlign(LangAS::Default);
+                        : CGM.getTarget().getPointerAlign(AS);
 
   VTable = CGM.CreateOrReplaceCXXRuntimeVariable(
       Name, VTableType, llvm::GlobalValue::ExternalLinkage,
@@ -3281,10 +3282,9 @@ ItaniumRTTIBuilder::GetAddrOfExternalRTTIDescriptor(QualType Ty) {
     // Note for the future: If we would ever like to do deferred emission of
     // RTTI, check if emitting vtables opportunistically need any adjustment.
 
-    GV = new llvm::GlobalVariable(CGM.getModule(), CGM.Int8PtrTy,
-                                  /*isConstant=*/true,
-                                  llvm::GlobalValue::ExternalLinkage, nullptr,
-                                  Name);
+    GV = new llvm::GlobalVariable(
+        CGM.getModule(), CGM.GlobalsInt8PtrTy,
+        /*isConstant=*/true, llvm::GlobalValue::ExternalLinkage, nullptr, Name);
     const CXXRecordDecl *RD = Ty->getAsCXXRecordDecl();
     CGM.setGVProperties(GV, RD);
     // Import the typeinfo symbol when all non-inline virtual methods are
@@ -3680,8 +3680,8 @@ void ItaniumRTTIBuilder::BuildVTablePointer(const Type *Ty) {
   if (CGM.getItaniumVTableContext().isRelativeLayout())
     VTable = CGM.getModule().getNamedAlias(VTableName);
   if (!VTable) {
-    llvm::Type *Ty = llvm::ArrayType::get(CGM.DefaultInt8PtrTy, 0);
-    VTable = CGM.CreateRuntimeVariable(Ty, VTableName);
+    llvm::Type *Ty = llvm::ArrayType::get(CGM.GlobalsInt8PtrTy, 0);
+    VTable = CGM.getModule().getOrInsertGlobal(VTableName, Ty);
   }
 
   CGM.setDSOLocal(cast<llvm::GlobalValue>(VTable->stripPointerCasts()));
@@ -3698,7 +3698,7 @@ void ItaniumRTTIBuilder::BuildVTablePointer(const Type *Ty) {
         llvm::ConstantExpr::getInBoundsGetElementPtr(CGM.Int8Ty, VTable, Eight);
   } else {
     llvm::Constant *Two = llvm::ConstantInt::get(PtrDiffTy, 2);
-    VTable = llvm::ConstantExpr::getInBoundsGetElementPtr(CGM.DefaultInt8PtrTy,
+    VTable = llvm::ConstantExpr::getInBoundsGetElementPtr(CGM.GlobalsInt8PtrTy,
                                                           VTable, Two);
   }
 
@@ -3835,7 +3835,7 @@ llvm::Constant *ItaniumRTTIBuilder::BuildTypeInfo(
         llvm::ConstantInt::get(CGM.Int64Ty, ((uint64_t)1) << 63);
     TypeNameField = llvm::ConstantExpr::getAdd(TypeNameField, flag);
     TypeNameField =
-        llvm::ConstantExpr::getIntToPtr(TypeNameField, CGM.Int8PtrTy);
+        llvm::ConstantExpr::getIntToPtr(TypeNameField, CGM.GlobalsInt8PtrTy);
   } else {
     TypeNameField = TypeName;
   }
@@ -3965,7 +3965,7 @@ llvm::Constant *ItaniumRTTIBuilder::BuildTypeInfo(
     GV->setComdat(M.getOrInsertComdat(GV->getName()));
 
   CharUnits Align = CGM.getContext().toCharUnitsFromBits(
-      CGM.getTarget().getPointerAlign(LangAS::Default));
+      CGM.getTarget().getPointerAlign(CGM.GetGlobalVarAddressSpace(nullptr)));
   GV->setAlignment(Align.getAsAlign());
 
   // The Itanium ABI specifies that type_info objects must be globally

clang/lib/Driver/Compilation.cpp

@@ -185,17 +185,22 @@ bool Compilation::CleanupFile(const char *File, bool IssueErrors) const {
   // able to remove), or non-regular files. Underlying tools may have
   // intentionally not overwritten them.
 
-  // Save the device code files(spv files) only if -fsycl-dump-device-code
-  // option is enabled.
+  // Save the device code files if -fsycl-dump-device-code option is enabled.
   if (TheDriver.isDumpDeviceCodeEnabled()) {
     Arg *DumpDeviceCodeArg =
         getArgs().getLastArg(options::OPT_fsycl_dump_device_code_EQ);
     std::string ExpectedDir =
         DumpDeviceCodeArg ? DumpDeviceCodeArg->getValue() : "";
     std::string ActualFile(File);
-    if (ActualFile.find(ExpectedDir) != std::string::npos &&
-        llvm::sys::path::extension(ActualFile).equals(".spv"))
-      return false;
+
+    if (ActualFile.find(ExpectedDir) != std::string::npos) {
+      // Save PTX files generated by LLVM NVPTX Back-End,
+      // when the nvptx*-nvidia-cuda is passed to -fsycl-targets.
+      if (DefaultToolChain.getTriple().isNVPTX())
+        return false;
+      if (llvm::sys::path::extension(ActualFile).equals(".spv"))
+        return false;
+    }
   }
 
   if (!llvm::sys::fs::can_write(File) || !llvm::sys::fs::is_regular_file(File))

clang/lib/Driver/ToolChains/MSVC.cpp

@@ -130,18 +130,23 @@ void visualstudio::Linker::ConstructJob(Compilation &C, const JobAction &JA,
     CmdArgs.push_back("-defaultlib:oldnames");
   }
 
-  if ((!C.getDriver().IsCLMode() && Args.hasArg(options::OPT_fsycl) &&
+  if ((Args.hasArg(options::OPT_fsycl) &&
        !Args.hasArg(options::OPT_nolibsycl)) ||
       Args.hasArg(options::OPT_fsycl_host_compiler_EQ)) {
     CmdArgs.push_back(Args.MakeArgString(std::string("-libpath:") +
                                          TC.getDriver().Dir + "/../lib"));
-    // When msvcrtd is added via --dependent-lib, we add the sycld
-    // equivalent.  Do not add the -defaultlib as it conflicts.
-    if (!isDependentLibAdded(Args, "msvcrtd")) {
+    if (!Args.hasArg(options::OPT__SLASH_MDd) &&
+        !isDependentLibAdded(Args, "msvcrtd")) {
       if (Args.hasArg(options::OPT_fpreview_breaking_changes))
         CmdArgs.push_back("-defaultlib:sycl" SYCL_MAJOR_VERSION "-preview.lib");
       else
         CmdArgs.push_back("-defaultlib:sycl" SYCL_MAJOR_VERSION ".lib");
+    } else {
+      if (Args.hasArg(options::OPT_fpreview_breaking_changes))
+        CmdArgs.push_back("-defaultlib:sycl" SYCL_MAJOR_VERSION
+                          "-previewd.lib");
+      else
+        CmdArgs.push_back("-defaultlib:sycl" SYCL_MAJOR_VERSION "d.lib");
     }
     CmdArgs.push_back("-defaultlib:sycl-devicelib-host.lib");
   }

clang/lib/Driver/ToolChains/SYCL.cpp

@@ -111,6 +111,26 @@ void SYCL::constructLLVMForeachCommand(Compilation &C, const JobAction &JA,
         C.getArgs().MakeArgString("--out-dir=" + OutputDirName));
   }
 
+  // If fsycl-dump-device-code is passed, put the PTX files
+  // into the path provided in fsycl-dump-device-code.
+  if (T->getToolChain().getTriple().isNVPTX() &&
+      C.getDriver().isDumpDeviceCodeEnabled() && Ext.equals("s")) {
+    SmallString<128> OutputDir;
+
+    Arg *DumpDeviceCodeArg =
+        C.getArgs().getLastArg(options::OPT_fsycl_dump_device_code_EQ);
+
+    OutputDir = (DumpDeviceCodeArg ? DumpDeviceCodeArg->getValue() : "");
+
+    // If the output directory path is empty, put the PTX files in the
+    // current directory.
+    if (OutputDir.empty())
+      llvm::sys::path::native(OutputDir = "./");
+    else
+      OutputDir.append(llvm::sys::path::get_separator());
+    ForeachArgs.push_back(C.getArgs().MakeArgString("--out-dir=" + OutputDir));
+  }
+
   ForeachArgs.push_back(C.getArgs().MakeArgString("--"));
   ForeachArgs.push_back(
       C.getArgs().MakeArgString(InputCommand->getExecutable()));