Merge branch 'urlog' of github.com:mateuszpn/llvm into urlog

Author: mateuszpn

.github/CODEOWNERS

@@ -56,6 +56,7 @@ sycl/source/detail/posix_ur.cpp @intel/unified-runtime-reviewers
 sycl/source/detail/ur.cpp @intel/unified-runtime-reviewers
 sycl/source/detail/windows_ur.cpp @intel/unified-runtime-reviewers
 sycl/test-e2e/Adapters/ @intel/unified-runtime-reviewers
+sycl/test-e2e/Adapters/level_zero/ @intel/unified-runtime-reviewers-level-zero
 
 # Win Proxy Loader
 sycl/ur_win_proxy_loader @intel/llvm-reviewers-runtime
@@ -202,9 +203,11 @@ llvm/include/llvm/Transforms/Instrumentation/AddressSanitizer.h @intel/dpcpp-san
 llvm/include/llvm/Transforms/Instrumentation/AddressSanitizerCommon.h @intel/dpcpp-sanitizers-review
 llvm/include/llvm/Transforms/Instrumentation/AddressSanitizerOptions.h @intel/dpcpp-sanitizers-review
 llvm/include/llvm/Transforms/Instrumentation/MemorySanitizer.h @intel/dpcpp-sanitizers-review
+llvm/include/llvm/Transforms/Instrumentation/SPIRVSanitizerCommonUtils.h @intel/dpcpp-sanitizers-review
 llvm/include/llvm/Transforms/Instrumentation/ThreadSanitizer.h @intel/dpcpp-sanitizers-review
 llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp @intel/dpcpp-sanitizers-review
 llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp @intel/dpcpp-sanitizers-review
+llvm/lib/Transforms/Instrumentation/SPIRVSanitizerCommonUtils.cpp @intel/dpcpp-sanitizers-review
 llvm/lib/Transforms/Instrumentation/ThreadSanitizer.cpp @intel/dpcpp-sanitizers-review
 llvm/test/Instrumentation/AddressSanitizer/ @intel/dpcpp-sanitizers-review
 llvm/test/Instrumentation/MemorySanitizer/ @intel/dpcpp-sanitizers-review

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

@@ -209,12 +209,12 @@ jobs:
         rm e2econf_files.txt
 
     - name: Pack E2E test binaries
-      if: ${{ always() && inputs.tests_selector == 'e2e' && inputs.e2e_testing_mode == 'build-only' }}
+      if: ${{ always() && !cancelled() && inputs.tests_selector == 'e2e' && inputs.e2e_testing_mode == 'build-only' }}
       shell: bash
       run: |
         tar -czf e2e_bin.tar.gz -C build-e2e .
     - name: Upload E2E test binaries
-      if: ${{ always() && inputs.tests_selector == 'e2e' && inputs.e2e_testing_mode == 'build-only' }}
+      if: ${{ always() && !cancelled() && inputs.tests_selector == 'e2e' && inputs.e2e_testing_mode == 'build-only' }}
       uses: actions/upload-artifact@v4
       with:
         name: ${{ inputs.e2e_binaries_artifact }}

clang/include/clang/Basic/CodeGenOptions.def

@@ -480,6 +480,11 @@ CODEGENOPT(DisableSYCLEarlyOpts, 1, 0)
 /// which do not contain "user" code.
 CODEGENOPT(OptimizeSYCLFramework, 1, 0)
 
+/// Whether to use alloca address space for `sret` arguments.
+/// TODO: This option can be removed once a fix goes in that can
+/// work with the community changes for using the alloca address space.
+CODEGENOPT(UseAllocaASForSrets, 1, 0)
+
 /// Turn on fp64 partial emulation for kernels with only fp64 conversion
 /// operations and no fp64 computation operations (requires Intel GPU backend
 /// supporting fp64 partial emulation)

clang/include/clang/Driver/Options.td

@@ -8830,6 +8830,13 @@ def fsycl_is_native_cpu : Flag<["-"], "fsycl-is-native-cpu">,
   HelpText<"Perform device compilation for Native CPU.">,
   Visibility<[CC1Option]>,
   MarshallingInfoFlag<LangOpts<"SYCLIsNativeCPU">>;
+// TODO: This option can be removed once a fix goes in that can
+// work with the community changes for using the alloca address space.
+defm offload_use_alloca_addrspace_for_srets : BoolFOption<"offload-use-alloca-addrspace-for-srets",
+  CodeGenOpts<"UseAllocaASForSrets">,
+  DefaultTrue,
+  PosFlag<SetTrue, [], [CC1Option], "Use alloca address space for sret arguments for offloading targets">,
+  NegFlag<SetFalse>>;
 
 } // let Visibility = [CC1Option]
 

clang/lib/Basic/Targets/NativeCPU.cpp

@@ -60,50 +60,12 @@ NativeCPUTargetInfo::NativeCPUTargetInfo(const llvm::Triple &,
   }());
   if (HostTriple.getArch() != llvm::Triple::UnknownArch) {
     HostTarget = AllocateTarget(HostTriple, Opts);
-
-    // Copy properties from host target.
-    BoolWidth = HostTarget->getBoolWidth();
-    BoolAlign = HostTarget->getBoolAlign();
-    IntWidth = HostTarget->getIntWidth();
-    IntAlign = HostTarget->getIntAlign();
-    HalfWidth = HostTarget->getHalfWidth();
-    HalfAlign = HostTarget->getHalfAlign();
-    FloatWidth = HostTarget->getFloatWidth();
-    FloatAlign = HostTarget->getFloatAlign();
-    DoubleWidth = HostTarget->getDoubleWidth();
-    DoubleAlign = HostTarget->getDoubleAlign();
-    LongWidth = HostTarget->getLongWidth();
-    LongAlign = HostTarget->getLongAlign();
-    LongLongWidth = HostTarget->getLongLongWidth();
-    LongLongAlign = HostTarget->getLongLongAlign();
-    PointerWidth = HostTarget->getPointerWidth(LangAS::Default);
-    PointerAlign = HostTarget->getPointerAlign(LangAS::Default);
-    MinGlobalAlign = HostTarget->getMinGlobalAlign(/*TypeSize=*/0,
-                                                   /*HasNonWeakDef=*/true);
-    NewAlign = HostTarget->getNewAlign();
-    DefaultAlignForAttributeAligned =
-        HostTarget->getDefaultAlignForAttributeAligned();
-    SizeType = HostTarget->getSizeType();
-    PtrDiffType = HostTarget->getPtrDiffType(LangAS::Default);
-    IntMaxType = HostTarget->getIntMaxType();
-    WCharType = HostTarget->getWCharType();
-    WIntType = HostTarget->getWIntType();
-    Char16Type = HostTarget->getChar16Type();
-    Char32Type = HostTarget->getChar32Type();
-    Int64Type = HostTarget->getInt64Type();
-    SigAtomicType = HostTarget->getSigAtomicType();
-    ProcessIDType = HostTarget->getProcessIDType();
-
-    UseBitFieldTypeAlignment = HostTarget->useBitFieldTypeAlignment();
-    UseZeroLengthBitfieldAlignment =
-        HostTarget->useZeroLengthBitfieldAlignment();
-    UseExplicitBitFieldAlignment = HostTarget->useExplicitBitFieldAlignment();
-    ZeroLengthBitfieldBoundary = HostTarget->getZeroLengthBitfieldBoundary();
-
-    // This is a bit of a lie, but it controls __GCC_ATOMIC_XXX_LOCK_FREE, and
-    // we need those macros to be identical on host and device, because (among
-    // other things) they affect which standard library classes are defined,
-    // and we need all classes to be defined on both the host and device.
-    MaxAtomicInlineWidth = HostTarget->getMaxAtomicInlineWidth();
+    copyAuxTarget(&*HostTarget);
   }
 }
+
+void NativeCPUTargetInfo::setAuxTarget(const TargetInfo *Aux) {
+  assert(Aux && "Cannot invoke setAuxTarget without a valid auxiliary target!");
+  copyAuxTarget(Aux);
+  getTargetOpts() = Aux->getTargetOpts();
+}

clang/lib/Basic/Targets/NativeCPU.h

@@ -57,6 +57,8 @@ class LLVM_LIBRARY_VISIBILITY NativeCPUTargetInfo final : public TargetInfo {
   }
 
 protected:
+  void setAuxTarget(const TargetInfo *Aux) override;
+
   ArrayRef<const char *> getGCCRegNames() const override { return {}; }
 
   ArrayRef<TargetInfo::GCCRegAlias> getGCCRegAliases() const override {

clang/lib/CodeGen/ABIInfoImpl.cpp

@@ -21,10 +21,16 @@ ABIArgInfo DefaultABIInfo::classifyArgumentType(QualType Ty) const {
     // Records with non-trivial destructors/copy-constructors should not be
     // passed by value.
     if (CGCXXABI::RecordArgABI RAA = getRecordArgABI(Ty, getCXXABI()))
-      return getNaturalAlignIndirect(Ty, getDataLayout().getAllocaAddrSpace(),
+      return getNaturalAlignIndirect(Ty,
+                                     getCodeGenOpts().UseAllocaASForSrets
+                                         ? getDataLayout().getAllocaAddrSpace()
+                                         : CGT.getTargetAddressSpace(Ty),
                                      RAA == CGCXXABI::RAA_DirectInMemory);
 
-    return getNaturalAlignIndirect(Ty, getDataLayout().getAllocaAddrSpace());
+    return getNaturalAlignIndirect(Ty,
+                                   getCodeGenOpts().UseAllocaASForSrets
+                                       ? getDataLayout().getAllocaAddrSpace()
+                                       : CGT.getTargetAddressSpace(Ty));
   }
 
   // Treat an enum type as its underlying type.
@@ -37,7 +43,10 @@ ABIArgInfo DefaultABIInfo::classifyArgumentType(QualType Ty) const {
         Context.getTypeSize(Context.getTargetInfo().hasInt128Type()
                                 ? Context.Int128Ty
                                 : Context.LongLongTy))
-      return getNaturalAlignIndirect(Ty, getDataLayout().getAllocaAddrSpace());
+      return getNaturalAlignIndirect(Ty,
+                                     getCodeGenOpts().UseAllocaASForSrets
+                                         ? getDataLayout().getAllocaAddrSpace()
+                                         : CGT.getTargetAddressSpace(Ty));
 
   return (isPromotableIntegerTypeForABI(Ty)
               ? ABIArgInfo::getExtend(Ty, CGT.ConvertType(Ty))
@@ -49,7 +58,10 @@ ABIArgInfo DefaultABIInfo::classifyReturnType(QualType RetTy) const {
     return ABIArgInfo::getIgnore();
 
   if (isAggregateTypeForABI(RetTy))
-    return getNaturalAlignIndirect(RetTy, getDataLayout().getAllocaAddrSpace());
+    return getNaturalAlignIndirect(RetTy,
+                                   getCodeGenOpts().UseAllocaASForSrets
+                                       ? getDataLayout().getAllocaAddrSpace()
+                                       : CGT.getTargetAddressSpace(RetTy));
 
   // Treat an enum type as its underlying type.
   if (const EnumType *EnumTy = RetTy->getAs<EnumType>())
@@ -61,7 +73,9 @@ ABIArgInfo DefaultABIInfo::classifyReturnType(QualType RetTy) const {
                                      ? getContext().Int128Ty
                                      : getContext().LongLongTy))
       return getNaturalAlignIndirect(RetTy,
-                                     getDataLayout().getAllocaAddrSpace());
+                                     getCodeGenOpts().UseAllocaASForSrets
+                                         ? getDataLayout().getAllocaAddrSpace()
+                                         : CGT.getTargetAddressSpace(RetTy));
 
   return (isPromotableIntegerTypeForABI(RetTy) ? ABIArgInfo::getExtend(RetTy)
                                                : ABIArgInfo::getDirect());
@@ -122,14 +136,16 @@ CGCXXABI::RecordArgABI CodeGen::getRecordArgABI(QualType T, CGCXXABI &CXXABI) {
 }
 
 bool CodeGen::classifyReturnType(const CGCXXABI &CXXABI, CGFunctionInfo &FI,
-                                 const ABIInfo &Info) {
+                                 const ABIInfo &Info, CodeGenTypes &CGT) {
   QualType Ty = FI.getReturnType();
 
   if (const auto *RT = Ty->getAs<RecordType>())
     if (!isa<CXXRecordDecl>(RT->getDecl()) &&
         !RT->getDecl()->canPassInRegisters()) {
       FI.getReturnInfo() = Info.getNaturalAlignIndirect(
-          Ty, Info.getDataLayout().getAllocaAddrSpace());
+          Ty, Info.getCodeGenOpts().UseAllocaASForSrets
+                  ? Info.getDataLayout().getAllocaAddrSpace()
+                  : CGT.getTargetAddressSpace(Ty));
       return true;
     }
 

clang/lib/CodeGen/ABIInfoImpl.h

@@ -46,7 +46,7 @@ CGCXXABI::RecordArgABI getRecordArgABI(const RecordType *RT, CGCXXABI &CXXABI);
 CGCXXABI::RecordArgABI getRecordArgABI(QualType T, CGCXXABI &CXXABI);
 
 bool classifyReturnType(const CGCXXABI &CXXABI, CGFunctionInfo &FI,
-                        const ABIInfo &Info);
+                        const ABIInfo &Info, CodeGenTypes &CGT);
 
 /// Pass transparent unions as if they were the type of the first element. Sema
 /// should ensure that all elements of the union have the same "machine type".

clang/lib/CodeGen/CGCall.cpp

@@ -1719,8 +1719,12 @@ CodeGenTypes::GetFunctionType(const CGFunctionInfo &FI) {
 
   // Add type for sret argument.
   if (IRFunctionArgs.hasSRetArg()) {
-    ArgTypes[IRFunctionArgs.getSRetArgNo()] = llvm::PointerType::get(
-        getLLVMContext(), FI.getReturnInfo().getIndirectAddrSpace());
+    QualType Ret = FI.getReturnType();
+    unsigned AddressSpace = CGM.getCodeGenOpts().UseAllocaASForSrets
+                                ? FI.getReturnInfo().getIndirectAddrSpace()
+                                : CGM.getTypes().getTargetAddressSpace(Ret);
+    ArgTypes[IRFunctionArgs.getSRetArgNo()] =
+        llvm::PointerType::get(getLLVMContext(), AddressSpace);
   }
 
   // Add type for inalloca argument.
@@ -5309,6 +5313,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
   // If the call returns a temporary with struct return, create a temporary
   // alloca to hold the result, unless one is given to us.
   Address SRetPtr = Address::invalid();
+  RawAddress SRetAlloca = RawAddress::invalid();
   llvm::Value *UnusedReturnSizePtr = nullptr;
   if (RetAI.isIndirect() || RetAI.isInAlloca() || RetAI.isCoerceAndExpand()) {
     // For virtual function pointer thunks and musttail calls, we must always
@@ -5322,19 +5327,27 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
     } else if (!ReturnValue.isNull()) {
       SRetPtr = ReturnValue.getAddress();
     } else {
-      SRetPtr = CreateMemTempWithoutCast(RetTy, "tmp");
+      SRetPtr = CGM.getCodeGenOpts().UseAllocaASForSrets
+                    ? CreateMemTempWithoutCast(RetTy, "tmp")
+                    : CreateMemTemp(RetTy, "tmp", &SRetAlloca);
       if (HaveInsertPoint() && ReturnValue.isUnused()) {
         llvm::TypeSize size =
             CGM.getDataLayout().getTypeAllocSize(ConvertTypeForMem(RetTy));
-        UnusedReturnSizePtr = EmitLifetimeStart(size, SRetPtr.getBasePointer());
+        if (CGM.getCodeGenOpts().UseAllocaASForSrets)
+          UnusedReturnSizePtr =
+              EmitLifetimeStart(size, SRetPtr.getBasePointer());
+        else
+          UnusedReturnSizePtr =
+              EmitLifetimeStart(size, SRetAlloca.getPointer());
       }
     }
     if (IRFunctionArgs.hasSRetArg()) {
       // A mismatch between the allocated return value's AS and the target's
       // chosen IndirectAS can happen e.g. when passing the this pointer through
       // a chain involving stores to / loads from the DefaultAS; we address this
       // here, symmetrically with the handling we have for normal pointer args.
-      if (SRetPtr.getAddressSpace() != RetAI.getIndirectAddrSpace()) {
+      if (CGM.getCodeGenOpts().UseAllocaASForSrets &&
+          (SRetPtr.getAddressSpace() != RetAI.getIndirectAddrSpace())) {
         llvm::Value *V = SRetPtr.getBasePointer();
         LangAS SAS = getLangASFromTargetAS(SRetPtr.getAddressSpace());
         LangAS DAS = getLangASFromTargetAS(RetAI.getIndirectAddrSpace());
@@ -5916,9 +5929,14 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
   // can't depend on being inside of an ExprWithCleanups, so we need to manually
   // pop this cleanup later on. Being eager about this is OK, since this
   // temporary is 'invisible' outside of the callee.
-  if (UnusedReturnSizePtr)
-    pushFullExprCleanup<CallLifetimeEnd>(NormalEHLifetimeMarker, SRetPtr,
-                                         UnusedReturnSizePtr);
+  if (UnusedReturnSizePtr) {
+    if (CGM.getCodeGenOpts().UseAllocaASForSrets)
+      pushFullExprCleanup<CallLifetimeEnd>(NormalEHLifetimeMarker, SRetPtr,
+                                           UnusedReturnSizePtr);
+    else
+      pushFullExprCleanup<CallLifetimeEnd>(NormalEHLifetimeMarker, SRetAlloca,
+                                           UnusedReturnSizePtr);
+  }
 
   llvm::BasicBlock *InvokeDest = CannotThrow ? nullptr : getInvokeDest();
 

clang/lib/CodeGen/ItaniumCXXABI.cpp

@@ -1349,10 +1349,14 @@ bool ItaniumCXXABI::classifyReturnType(CGFunctionInfo &FI) const {
 
   // If C++ prohibits us from making a copy, return by address.
   if (!RD->canPassInRegisters()) {
-    auto Align = CGM.getContext().getTypeAlignInChars(FI.getReturnType());
-    FI.getReturnInfo() = ABIArgInfo::getIndirect(
-        Align, /*AddrSpace=*/CGM.getDataLayout().getAllocaAddrSpace(),
-        /*ByVal=*/false);
+    QualType Ret = FI.getReturnType();
+    auto Align = CGM.getContext().getTypeAlignInChars(Ret);
+    unsigned AddressSpace = CGM.getCodeGenOpts().UseAllocaASForSrets
+                                ? CGM.getDataLayout().getAllocaAddrSpace()
+                                : CGM.getTypes().getTargetAddressSpace(Ret);
+    FI.getReturnInfo() =
+        ABIArgInfo::getIndirect(Align, /*AddrSpace=*/AddressSpace,
+                                /*ByVal=*/false);
     return true;
   }
   return false;