From: Alexey Bader Date: Fri, 20 Dec 2019 09:05:42 +0000 (+0300) Subject: [SYCL] Implement SYCL address space attributes handling X-Git-Tag: llvmorg-14-init~8461 X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=7818906ca134775edffb77857f436359d3a50b90;p=platform%2Fupstream%2Fllvm.git [SYCL] Implement SYCL address space attributes handling Default address space (applies when no explicit address space was specified) maps to generic (4) address space. Added SYCL named address spaces `sycl_global`, `sycl_local` and `sycl_private` defined as sub-sets of the default address space. Static variables without address space now reside in global address space when compile for SPIR target, unless they have an explicit address space qualifier in source code. Differential Revision: https://reviews.llvm.org/D89909 --- diff --git a/clang/include/clang/AST/Type.h b/clang/include/clang/AST/Type.h index 530b3ce..84de51a 100644 --- a/clang/include/clang/AST/Type.h +++ b/clang/include/clang/AST/Type.h @@ -488,7 +488,11 @@ public: B == LangAS::opencl_global_host)) || // Consider pointer size address spaces to be equivalent to default. ((isPtrSizeAddressSpace(A) || A == LangAS::Default) && - (isPtrSizeAddressSpace(B) || B == LangAS::Default)); + (isPtrSizeAddressSpace(B) || B == LangAS::Default)) || + // Default is a superset of SYCL address spaces. + (A == LangAS::Default && + (B == LangAS::sycl_private || B == LangAS::sycl_local || + B == LangAS::sycl_global)); } /// Returns true if the address space in these qualifiers is equal to or diff --git a/clang/include/clang/Basic/AddressSpaces.h b/clang/include/clang/Basic/AddressSpaces.h index a9db52d..5fa0312 100644 --- a/clang/include/clang/Basic/AddressSpaces.h +++ b/clang/include/clang/Basic/AddressSpaces.h @@ -44,6 +44,11 @@ enum class LangAS : unsigned { cuda_constant, cuda_shared, + // SYCL specific address spaces. + sycl_global, + sycl_local, + sycl_private, + // Pointer size and extension address spaces. ptr32_sptr, ptr32_uptr, diff --git a/clang/include/clang/Sema/ParsedAttr.h b/clang/include/clang/Sema/ParsedAttr.h index e508b46..3479258 100644 --- a/clang/include/clang/Sema/ParsedAttr.h +++ b/clang/include/clang/Sema/ParsedAttr.h @@ -628,8 +628,8 @@ public: /// a Spelling enumeration, the value UINT_MAX is returned. unsigned getSemanticSpelling() const; - /// If this is an OpenCL addr space attribute returns its representation - /// in LangAS, otherwise returns default addr space. + /// If this is an OpenCL address space attribute returns its representation + /// in LangAS, otherwise returns default address space. LangAS asOpenCLLangAS() const { switch (getParsedKind()) { case ParsedAttr::AT_OpenCLConstantAddressSpace: @@ -651,6 +651,22 @@ public: } } + /// If this is an OpenCL address space attribute returns its SYCL + /// representation in LangAS, otherwise returns default address space. + LangAS asSYCLLangAS() const { + switch (getKind()) { + case ParsedAttr::AT_OpenCLGlobalAddressSpace: + return LangAS::sycl_global; + case ParsedAttr::AT_OpenCLLocalAddressSpace: + return LangAS::sycl_local; + case ParsedAttr::AT_OpenCLPrivateAddressSpace: + return LangAS::sycl_private; + case ParsedAttr::AT_OpenCLGenericAddressSpace: + default: + return LangAS::Default; + } + } + AttributeCommonInfo::Kind getKind() const { return AttributeCommonInfo::Kind(Info.AttrKind); } diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp index aefb2ad..85e5a61 100644 --- a/clang/lib/AST/ASTContext.cpp +++ b/clang/lib/AST/ASTContext.cpp @@ -931,6 +931,9 @@ static const LangASMap *getAddressSpaceMap(const TargetInfo &T, 7, // cuda_device 8, // cuda_constant 9, // cuda_shared + 1, // sycl_global + 3, // sycl_local + 0, // sycl_private 10, // ptr32_sptr 11, // ptr32_uptr 12 // ptr64 diff --git a/clang/lib/AST/ItaniumMangle.cpp b/clang/lib/AST/ItaniumMangle.cpp index 6415fca..79fec68 100644 --- a/clang/lib/AST/ItaniumMangle.cpp +++ b/clang/lib/AST/ItaniumMangle.cpp @@ -2462,7 +2462,8 @@ void CXXNameMangler::mangleQualifiers(Qualifiers Quals, const DependentAddressSp if (Context.getASTContext().addressSpaceMapManglingFor(AS)) { // ::= "AS" unsigned TargetAS = Context.getASTContext().getTargetAddressSpace(AS); - if (TargetAS != 0) + if (TargetAS != 0 || + Context.getASTContext().getTargetAddressSpace(LangAS::Default) != 0) ASString = "AS" + llvm::utostr(TargetAS); } else { switch (AS) { @@ -2491,6 +2492,16 @@ void CXXNameMangler::mangleQualifiers(Qualifiers Quals, const DependentAddressSp case LangAS::opencl_generic: ASString = "CLgeneric"; break; + // ::= "SY" [ "global" | "local" | "private" ] + case LangAS::sycl_global: + ASString = "SYglobal"; + break; + case LangAS::sycl_local: + ASString = "SYlocal"; + break; + case LangAS::sycl_private: + ASString = "SYprivate"; + break; // ::= "CU" [ "device" | "constant" | "shared" ] case LangAS::cuda_device: ASString = "CUdevice"; diff --git a/clang/lib/AST/TypePrinter.cpp b/clang/lib/AST/TypePrinter.cpp index 17e9b3f..9000976f 100644 --- a/clang/lib/AST/TypePrinter.cpp +++ b/clang/lib/AST/TypePrinter.cpp @@ -2111,10 +2111,13 @@ std::string Qualifiers::getAddrSpaceAsString(LangAS AS) { case LangAS::Default: return ""; case LangAS::opencl_global: + case LangAS::sycl_global: return "__global"; case LangAS::opencl_local: + case LangAS::sycl_local: return "__local"; case LangAS::opencl_private: + case LangAS::sycl_private: return "__private"; case LangAS::opencl_constant: return "__constant"; diff --git a/clang/lib/Basic/Targets/AMDGPU.cpp b/clang/lib/Basic/Targets/AMDGPU.cpp index 75115db..6689fba 100644 --- a/clang/lib/Basic/Targets/AMDGPU.cpp +++ b/clang/lib/Basic/Targets/AMDGPU.cpp @@ -51,6 +51,9 @@ const LangASMap AMDGPUTargetInfo::AMDGPUDefIsGenMap = { Global, // cuda_device Constant, // cuda_constant Local, // cuda_shared + Global, // sycl_global + Local, // sycl_local + Private, // sycl_private Generic, // ptr32_sptr Generic, // ptr32_uptr Generic // ptr64 @@ -68,6 +71,10 @@ const LangASMap AMDGPUTargetInfo::AMDGPUDefIsPrivMap = { Global, // cuda_device Constant, // cuda_constant Local, // cuda_shared + // SYCL address space values for this map are dummy + Generic, // sycl_global + Generic, // sycl_local + Generic, // sycl_private Generic, // ptr32_sptr Generic, // ptr32_uptr Generic // ptr64 diff --git a/clang/lib/Basic/Targets/NVPTX.h b/clang/lib/Basic/Targets/NVPTX.h index 038dec4..9e80473 100644 --- a/clang/lib/Basic/Targets/NVPTX.h +++ b/clang/lib/Basic/Targets/NVPTX.h @@ -35,6 +35,9 @@ static const unsigned NVPTXAddrSpaceMap[] = { 1, // cuda_device 4, // cuda_constant 3, // cuda_shared + 1, // sycl_global + 3, // sycl_local + 0, // sycl_private 0, // ptr32_sptr 0, // ptr32_uptr 0 // ptr64 diff --git a/clang/lib/Basic/Targets/SPIR.h b/clang/lib/Basic/Targets/SPIR.h index a2e8126..6d19016 100644 --- a/clang/lib/Basic/Targets/SPIR.h +++ b/clang/lib/Basic/Targets/SPIR.h @@ -21,7 +21,7 @@ namespace clang { namespace targets { -static const unsigned SPIRAddrSpaceMap[] = { +static const unsigned SPIRDefIsPrivMap[] = { 0, // Default 1, // opencl_global 3, // opencl_local @@ -33,6 +33,31 @@ static const unsigned SPIRAddrSpaceMap[] = { 0, // cuda_device 0, // cuda_constant 0, // cuda_shared + // SYCL address space values for this map are dummy + 0, // sycl_global + 0, // sycl_local + 0, // sycl_private + 0, // ptr32_sptr + 0, // ptr32_uptr + 0 // ptr64 +}; + +static const unsigned SPIRDefIsGenMap[] = { + 4, // Default + // OpenCL address space values for this map are dummy and they can't be used + 0, // opencl_global + 0, // opencl_local + 0, // opencl_constant + 0, // opencl_private + 0, // opencl_generic + 0, // opencl_global_device + 0, // opencl_global_host + 0, // cuda_device + 0, // cuda_constant + 0, // cuda_shared + 1, // sycl_global + 3, // sycl_local + 0, // sycl_private 0, // ptr32_sptr 0, // ptr32_uptr 0 // ptr64 @@ -49,7 +74,7 @@ public: TLSSupported = false; VLASupported = false; LongWidth = LongAlign = 64; - AddrSpaceMap = &SPIRAddrSpaceMap; + AddrSpaceMap = &SPIRDefIsPrivMap; UseAddrSpaceMapMangling = true; HasLegalHalfType = true; HasFloat16 = true; @@ -97,6 +122,22 @@ public: return CC_SpirFunction; } + void setAddressSpaceMap(bool DefaultIsGeneric) { + AddrSpaceMap = DefaultIsGeneric ? &SPIRDefIsGenMap : &SPIRDefIsPrivMap; + } + + void adjust(LangOptions &Opts) override { + TargetInfo::adjust(Opts); + // FIXME: SYCL specification considers unannotated pointers and references + // to be pointing to the generic address space. See section 5.9.3 of + // SYCL 2020 specification. + // Currently, there is no way of representing SYCL's default address space + // language semantic along with the semantics of embedded C's default + // address space in the same address space map. Hence the map needs to be + // reset to allow mapping to the desired value of 'Default' entry for SYCL. + setAddressSpaceMap(/*DefaultIsGeneric=*/Opts.SYCLIsDevice); + } + void setSupportedOpenCLOpts() override { // Assume all OpenCL extensions and optional core features are supported // for SPIR since it is a generic target. @@ -107,6 +148,7 @@ public: bool hasInt128Type() const override { return false; } }; + class LLVM_LIBRARY_VISIBILITY SPIR32TargetInfo : public SPIRTargetInfo { public: SPIR32TargetInfo(const llvm::Triple &Triple, const TargetOptions &Opts) diff --git a/clang/lib/Basic/Targets/TCE.h b/clang/lib/Basic/Targets/TCE.h index 445fe4f..5a122be 100644 --- a/clang/lib/Basic/Targets/TCE.h +++ b/clang/lib/Basic/Targets/TCE.h @@ -42,6 +42,9 @@ static const unsigned TCEOpenCLAddrSpaceMap[] = { 0, // cuda_device 0, // cuda_constant 0, // cuda_shared + 3, // sycl_global + 4, // sycl_local + 0, // sycl_private 0, // ptr32_sptr 0, // ptr32_uptr 0, // ptr64 diff --git a/clang/lib/Basic/Targets/X86.h b/clang/lib/Basic/Targets/X86.h index 91a365c..0858099 100644 --- a/clang/lib/Basic/Targets/X86.h +++ b/clang/lib/Basic/Targets/X86.h @@ -35,6 +35,9 @@ static const unsigned X86AddrSpaceMap[] = { 0, // cuda_device 0, // cuda_constant 0, // cuda_shared + 0, // sycl_global + 0, // sycl_local + 0, // sycl_private 270, // ptr32_sptr 271, // ptr32_uptr 272 // ptr64 diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp index fda78f8..5bab185 100644 --- a/clang/lib/CodeGen/CGDecl.cpp +++ b/clang/lib/CodeGen/CGDecl.cpp @@ -1122,7 +1122,7 @@ Address CodeGenModule::createUnnamedGlobalFrom(const VarDecl &D, bool isConstant = true; llvm::GlobalVariable *InsertBefore = nullptr; unsigned AS = - getContext().getTargetAddressSpace(getStringLiteralAddressSpace()); + getContext().getTargetAddressSpace(GetGlobalConstantAddressSpace()); std::string Name; if (D.hasGlobalStorage()) Name = getMangledName(&D).str() + ".const"; diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp index a35bd7d..b988659 100644 --- a/clang/lib/CodeGen/CGExpr.cpp +++ b/clang/lib/CodeGen/CGExpr.cpp @@ -404,24 +404,22 @@ static Address createReferenceTemporary(CodeGenFunction &CGF, (Ty->isArrayType() || Ty->isRecordType()) && CGF.CGM.isTypeConstant(Ty, true)) if (auto Init = ConstantEmitter(CGF).tryEmitAbstract(Inner, Ty)) { - if (auto AddrSpace = CGF.getTarget().getConstantAddressSpace()) { - auto AS = AddrSpace.getValue(); - auto *GV = new llvm::GlobalVariable( - CGF.CGM.getModule(), Init->getType(), /*isConstant=*/true, - llvm::GlobalValue::PrivateLinkage, Init, ".ref.tmp", nullptr, - llvm::GlobalValue::NotThreadLocal, - CGF.getContext().getTargetAddressSpace(AS)); - CharUnits alignment = CGF.getContext().getTypeAlignInChars(Ty); - GV->setAlignment(alignment.getAsAlign()); - llvm::Constant *C = GV; - if (AS != LangAS::Default) - C = TCG.performAddrSpaceCast( - CGF.CGM, GV, AS, LangAS::Default, - GV->getValueType()->getPointerTo( - CGF.getContext().getTargetAddressSpace(LangAS::Default))); - // FIXME: Should we put the new global into a COMDAT? - return Address(C, alignment); - } + auto AS = CGF.CGM.GetGlobalConstantAddressSpace(); + auto *GV = new llvm::GlobalVariable( + CGF.CGM.getModule(), Init->getType(), /*isConstant=*/true, + llvm::GlobalValue::PrivateLinkage, Init, ".ref.tmp", nullptr, + llvm::GlobalValue::NotThreadLocal, + CGF.getContext().getTargetAddressSpace(AS)); + CharUnits alignment = CGF.getContext().getTypeAlignInChars(Ty); + GV->setAlignment(alignment.getAsAlign()); + llvm::Constant *C = GV; + if (AS != LangAS::Default) + C = TCG.performAddrSpaceCast( + CGF.CGM, GV, AS, LangAS::Default, + GV->getValueType()->getPointerTo( + CGF.getContext().getTargetAddressSpace(LangAS::Default))); + // FIXME: Should we put the new global into a COMDAT? + return Address(C, alignment); } return CGF.CreateMemTemp(Ty, "ref.tmp", Alloca); } diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp index fd5975b..8d4a0e0 100644 --- a/clang/lib/CodeGen/CodeGenModule.cpp +++ b/clang/lib/CodeGen/CodeGenModule.cpp @@ -4116,6 +4116,10 @@ LangAS CodeGenModule::GetGlobalVarAddressSpace(const VarDecl *D) { return AddrSpace; } + if (LangOpts.SYCLIsDevice && + (!D || D->getType().getAddressSpace() == LangAS::Default)) + return LangAS::sycl_global; + if (LangOpts.CUDA && LangOpts.CUDAIsDevice) { if (D && D->hasAttr()) return LangAS::cuda_constant; @@ -4137,10 +4141,12 @@ LangAS CodeGenModule::GetGlobalVarAddressSpace(const VarDecl *D) { return getTargetCodeGenInfo().getGlobalVarAddressSpace(*this, D); } -LangAS CodeGenModule::getStringLiteralAddressSpace() const { +LangAS CodeGenModule::GetGlobalConstantAddressSpace() const { // OpenCL v1.2 s6.5.3: a string literal is in the constant address space. if (LangOpts.OpenCL) return LangAS::opencl_constant; + if (LangOpts.SYCLIsDevice) + return LangAS::sycl_global; if (auto AS = getTarget().getConstantAddressSpace()) return AS.getValue(); return LangAS::Default; @@ -4159,13 +4165,12 @@ castStringLiteralToDefaultAddressSpace(CodeGenModule &CGM, llvm::GlobalVariable *GV) { llvm::Constant *Cast = GV; if (!CGM.getLangOpts().OpenCL) { - if (auto AS = CGM.getTarget().getConstantAddressSpace()) { - if (AS != LangAS::Default) - Cast = CGM.getTargetCodeGenInfo().performAddrSpaceCast( - CGM, GV, AS.getValue(), LangAS::Default, - GV->getValueType()->getPointerTo( - CGM.getContext().getTargetAddressSpace(LangAS::Default))); - } + auto AS = CGM.GetGlobalConstantAddressSpace(); + if (AS != LangAS::Default) + Cast = CGM.getTargetCodeGenInfo().performAddrSpaceCast( + CGM, GV, AS, LangAS::Default, + GV->getValueType()->getPointerTo( + CGM.getContext().getTargetAddressSpace(LangAS::Default))); } return Cast; } @@ -5277,7 +5282,7 @@ GenerateStringLiteral(llvm::Constant *C, llvm::GlobalValue::LinkageTypes LT, CodeGenModule &CGM, StringRef GlobalName, CharUnits Alignment) { unsigned AddrSpace = CGM.getContext().getTargetAddressSpace( - CGM.getStringLiteralAddressSpace()); + CGM.GetGlobalConstantAddressSpace()); llvm::Module &M = CGM.getModule(); // Create a global variable for this string diff --git a/clang/lib/CodeGen/CodeGenModule.h b/clang/lib/CodeGen/CodeGenModule.h index 83230db..32d74c5 100644 --- a/clang/lib/CodeGen/CodeGenModule.h +++ b/clang/lib/CodeGen/CodeGenModule.h @@ -855,6 +855,13 @@ public: /// space, target-specific global or constant address space may be returned. LangAS GetGlobalVarAddressSpace(const VarDecl *D); + /// Return the AST address space of constant literal, which is used to emit + /// the constant literal as global variable in LLVM IR. + /// Note: This is not necessarily the address space of the constant literal + /// in AST. For address space agnostic language, e.g. C++, constant literal + /// in AST is always in default address space. + LangAS GetGlobalConstantAddressSpace() const; + /// Return the llvm::Constant for the address of the given global variable. /// If Ty is non-null and if the global doesn't exist, then it will be created /// with the specified type instead of whatever the normal requested type @@ -866,13 +873,6 @@ public: ForDefinition_t IsForDefinition = NotForDefinition); - /// Return the AST address space of string literal, which is used to emit - /// the string literal as global variable in LLVM IR. - /// Note: This is not necessarily the address space of the string literal - /// in AST. For address space agnostic language, e.g. C++, string literal - /// in AST is always in default address space. - LangAS getStringLiteralAddressSpace() const; - /// Return the address of the given function. If Ty is non-null, then this /// function will use the specified type if it has to create it. llvm::Constant *GetAddrOfFunction(GlobalDecl GD, llvm::Type *Ty = nullptr, diff --git a/clang/lib/CodeGen/TargetInfo.cpp b/clang/lib/CodeGen/TargetInfo.cpp index 3e0a269..52dce97 100644 --- a/clang/lib/CodeGen/TargetInfo.cpp +++ b/clang/lib/CodeGen/TargetInfo.cpp @@ -10068,6 +10068,12 @@ class SPIRTargetCodeGenInfo : public TargetCodeGenInfo { public: SPIRTargetCodeGenInfo(CodeGen::CodeGenTypes &CGT) : TargetCodeGenInfo(std::make_unique(CGT)) {} + + LangAS getASTAllocaAddressSpace() const override { + return getLangASFromTargetAS( + getABIInfo().getDataLayout().getAllocaAddrSpace()); + } + unsigned getOpenCLKernelCallingConv() const override; }; diff --git a/clang/lib/Sema/SemaType.cpp b/clang/lib/Sema/SemaType.cpp index ed0480e..9e826eb 100644 --- a/clang/lib/Sema/SemaType.cpp +++ b/clang/lib/Sema/SemaType.cpp @@ -6378,6 +6378,7 @@ static bool BuildAddressSpaceIndex(Sema &S, LangAS &ASIdx, llvm::APSInt max(addrSpace.getBitWidth()); max = Qualifiers::MaxAddressSpace - (unsigned)LangAS::FirstTargetAddressSpace; + if (addrSpace > max) { S.Diag(AttrLoc, diag::err_attribute_address_space_too_high) << (unsigned)max.getZExtValue() << AddrSpace->getSourceRange(); @@ -6493,7 +6494,9 @@ static void HandleAddressSpaceTypeAttribute(QualType &Type, Attr.setInvalid(); } else { // The keyword-based type attributes imply which address space to use. - ASIdx = Attr.asOpenCLLangAS(); + ASIdx = S.getLangOpts().SYCLIsDevice ? Attr.asSYCLLangAS() + : Attr.asOpenCLLangAS(); + if (ASIdx == LangAS::Default) llvm_unreachable("Invalid address space"); diff --git a/clang/test/CodeGenSYCL/address-space-conversions.cpp b/clang/test/CodeGenSYCL/address-space-conversions.cpp new file mode 100644 index 0000000..2da9963 --- /dev/null +++ b/clang/test/CodeGenSYCL/address-space-conversions.cpp @@ -0,0 +1,125 @@ +// RUN: %clang_cc1 -triple spir64 -fsycl-is-device -disable-llvm-passes -emit-llvm %s -o - | FileCheck %s +void bar(int &Data) {} +// CHECK-DAG: define{{.*}} spir_func void @[[RAW_REF:[a-zA-Z0-9_]+]](i32 addrspace(4)* align 4 dereferenceable(4) % +void bar2(int &Data) {} +// CHECK-DAG: define{{.*}} spir_func void @[[RAW_REF2:[a-zA-Z0-9_]+]](i32 addrspace(4)* align 4 dereferenceable(4) % +void bar(__attribute__((opencl_local)) int &Data) {} +// CHECK-DAG: define{{.*}} spir_func void [[LOC_REF:@[a-zA-Z0-9_]+]](i32 addrspace(3)* align 4 dereferenceable(4) % +void foo(int *Data) {} +// CHECK-DAG: define{{.*}} spir_func void @[[RAW_PTR:[a-zA-Z0-9_]+]](i32 addrspace(4)* % +void foo2(int *Data) {} +// CHECK-DAG: define{{.*}} spir_func void @[[RAW_PTR2:[a-zA-Z0-9_]+]](i32 addrspace(4)* % +void foo(__attribute__((opencl_local)) int *Data) {} +// CHECK-DAG: define{{.*}} spir_func void [[LOC_PTR:@[a-zA-Z0-9_]+]](i32 addrspace(3)* % + +template +void tmpl(T t) {} +// See Check Lines below. + +void usages() { + // CHECK-DAG: [[GLOB:%[a-zA-Z0-9]+]] = alloca i32 addrspace(1)* + // CHECK-DAG: [[GLOB]].ascast = addrspacecast i32 addrspace(1)** [[GLOB]] to i32 addrspace(1)* addrspace(4)* + __attribute__((opencl_global)) int *GLOB; + // CHECK-DAG: [[LOC:%[a-zA-Z0-9]+]] = alloca i32 addrspace(3)* + // CHECK-DAG: [[LOC]].ascast = addrspacecast i32 addrspace(3)** [[LOC]] to i32 addrspace(3)* addrspace(4)* + __attribute__((opencl_local)) int *LOC; + // CHECK-DAG: [[NoAS:%[a-zA-Z0-9]+]] = alloca i32 addrspace(4)* + // CHECK-DAG: [[NoAS]].ascast = addrspacecast i32 addrspace(4)** [[NoAS]] to i32 addrspace(4)* addrspace(4)* + int *NoAS; + // CHECK-DAG: [[PRIV:%[a-zA-Z0-9]+]] = alloca i32* + // CHECK-DAG: [[PRIV]].ascast = addrspacecast i32** [[PRIV]] to i32* addrspace(4)* + __attribute__((opencl_private)) int *PRIV; + + // Explicit conversions + // From names address spaces to default address space + // CHECK-DAG: [[GLOB_LOAD:%[a-zA-Z0-9]+]] = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)* [[GLOB]].ascast + // CHECK-DAG: [[GLOB_CAST:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(1)* [[GLOB_LOAD]] to i32 addrspace(4)* + // CHECK-DAG: store i32 addrspace(4)* [[GLOB_CAST]], i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + NoAS = (int *)GLOB; + // CHECK-DAG: [[LOC_LOAD:%[a-zA-Z0-9]+]] = load i32 addrspace(3)*, i32 addrspace(3)* addrspace(4)* [[LOC]].ascast + // CHECK-DAG: [[LOC_CAST:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(3)* [[LOC_LOAD]] to i32 addrspace(4)* + // CHECK-DAG: store i32 addrspace(4)* [[LOC_CAST]], i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + NoAS = (int *)LOC; + // CHECK-DAG: [[PRIV_LOAD:%[a-zA-Z0-9]+]] = load i32*, i32* addrspace(4)* [[PRIV]].ascast + // CHECK-DAG: [[PRIV_CAST:%[a-zA-Z0-9]+]] = addrspacecast i32* [[PRIV_LOAD]] to i32 addrspace(4)* + // CHECK-DAG: store i32 addrspace(4)* [[PRIV_CAST]], i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + NoAS = (int *)PRIV; + // From default address space to named address space + // CHECK-DAG: [[NoAS_LOAD:%[a-zA-Z0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + // CHECK-DAG: [[NoAS_CAST:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(4)* [[NoAS_LOAD]] to i32 addrspace(1)* + // CHECK-DAG: store i32 addrspace(1)* [[NoAS_CAST]], i32 addrspace(1)* addrspace(4)* [[GLOB]].ascast + GLOB = (__attribute__((opencl_global)) int *)NoAS; + // CHECK-DAG: [[NoAS_LOAD:%[a-zA-Z0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + // CHECK-DAG: [[NoAS_CAST:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(4)* [[NoAS_LOAD]] to i32 addrspace(3)* + // CHECK-DAG: store i32 addrspace(3)* [[NoAS_CAST]], i32 addrspace(3)* addrspace(4)* [[LOC]].ascast + LOC = (__attribute__((opencl_local)) int *)NoAS; + // CHECK-DAG: [[NoAS_LOAD:%[a-zA-Z0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + // CHECK-DAG: [[NoAS_CAST:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(4)* [[NoAS_LOAD]] to i32* + // CHECK-DAG: store i32* [[NoAS_CAST]], i32* addrspace(4)* [[PRIV]].ascast + PRIV = (__attribute__((opencl_private)) int *)NoAS; + + bar(*GLOB); + // CHECK-DAG: [[GLOB_LOAD:%[a-zA-Z0-9]+]] = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)* [[GLOB]].ascast + // CHECK-DAG: [[GLOB_CAST:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(1)* [[GLOB_LOAD]] to i32 addrspace(4)* + // CHECK-DAG: call spir_func void @[[RAW_REF]](i32 addrspace(4)* align 4 dereferenceable(4) [[GLOB_CAST]]) + bar2(*GLOB); + // CHECK-DAG: [[GLOB_LOAD2:%[a-zA-Z0-9]+]] = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)* [[GLOB]].ascast + // CHECK-DAG: [[GLOB_CAST2:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(1)* [[GLOB_LOAD2]] to i32 addrspace(4)* + // CHECK-DAG: call spir_func void @[[RAW_REF2]](i32 addrspace(4)* align 4 dereferenceable(4) [[GLOB_CAST2]]) + + bar(*LOC); + // CHECK-DAG: [[LOC_LOAD:%[a-zA-Z0-9]+]] = load i32 addrspace(3)*, i32 addrspace(3)* addrspace(4)* [[LOC]].ascast + // CHECK-DAG: call spir_func void [[LOC_REF]](i32 addrspace(3)* align 4 dereferenceable(4) [[LOC_LOAD]]) + bar2(*LOC); + // CHECK-DAG: [[LOC_LOAD2:%[a-zA-Z0-9]+]] = load i32 addrspace(3)*, i32 addrspace(3)* addrspace(4)* [[LOC]].ascast + // CHECK-DAG: [[LOC_CAST2:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(3)* [[LOC_LOAD2]] to i32 addrspace(4)* + // CHECK-DAG: call spir_func void @[[RAW_REF2]](i32 addrspace(4)* align 4 dereferenceable(4) [[LOC_CAST2]]) + + bar(*NoAS); + // CHECK-DAG: [[NoAS_LOAD:%[a-zA-Z0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + // CHECK-DAG: call spir_func void @[[RAW_REF]](i32 addrspace(4)* align 4 dereferenceable(4) [[NoAS_LOAD]]) + bar2(*NoAS); + // CHECK-DAG: [[NoAS_LOAD2:%[a-zA-Z0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + // CHECK-DAG: call spir_func void @[[RAW_REF2]](i32 addrspace(4)* align 4 dereferenceable(4) [[NoAS_LOAD2]]) + + foo(GLOB); + // CHECK-DAG: [[GLOB_LOAD3:%[a-zA-Z0-9]+]] = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)* [[GLOB]].ascast + // CHECK-DAG: [[GLOB_CAST3:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(1)* [[GLOB_LOAD3]] to i32 addrspace(4)* + // CHECK-DAG: call spir_func void @[[RAW_PTR]](i32 addrspace(4)* [[GLOB_CAST3]]) + foo2(GLOB); + // CHECK-DAG: [[GLOB_LOAD4:%[a-zA-Z0-9]+]] = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)* [[GLOB]].ascast + // CHECK-DAG: [[GLOB_CAST4:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(1)* [[GLOB_LOAD4]] to i32 addrspace(4)* + // CHECK-DAG: call spir_func void @[[RAW_PTR2]](i32 addrspace(4)* [[GLOB_CAST4]]) + foo(LOC); + // CHECK-DAG: [[LOC_LOAD3:%[a-zA-Z0-9]+]] = load i32 addrspace(3)*, i32 addrspace(3)* addrspace(4)* [[LOC]].ascast + // CHECK-DAG: call spir_func void [[LOC_PTR]](i32 addrspace(3)* [[LOC_LOAD3]]) + foo2(LOC); + // CHECK-DAG: [[LOC_LOAD4:%[a-zA-Z0-9]+]] = load i32 addrspace(3)*, i32 addrspace(3)* addrspace(4)* [[LOC]].ascast + // CHECK-DAG: [[LOC_CAST4:%[a-zA-Z0-9]+]] = addrspacecast i32 addrspace(3)* [[LOC_LOAD4]] to i32 addrspace(4)* + // CHECK-DAG: call spir_func void @[[RAW_PTR2]](i32 addrspace(4)* [[LOC_CAST4]]) + foo(NoAS); + // CHECK-DAG: [[NoAS_LOAD3:%[a-zA-Z0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + // CHECK-DAG: call spir_func void @[[RAW_PTR]](i32 addrspace(4)* [[NoAS_LOAD3]]) + foo2(NoAS); + // CHECK-DAG: [[NoAS_LOAD4:%[a-zA-Z0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + // CHECK-DAG: call spir_func void @[[RAW_PTR2]](i32 addrspace(4)* [[NoAS_LOAD4]]) + + // Ensure that we still get 3 different template instantiations. + tmpl(GLOB); + // CHECK-DAG: [[GLOB_LOAD4:%[a-zA-Z0-9]+]] = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)* [[GLOB]].ascast + // CHECK-DAG: call spir_func void @_Z4tmplIPU3AS1iEvT_(i32 addrspace(1)* [[GLOB_LOAD4]]) + tmpl(LOC); + // CHECK-DAG: [[LOC_LOAD5:%[a-zA-Z0-9]+]] = load i32 addrspace(3)*, i32 addrspace(3)* addrspace(4)* [[LOC]].ascast + // CHECK-DAG: call spir_func void @_Z4tmplIPU3AS3iEvT_(i32 addrspace(3)* [[LOC_LOAD5]]) + tmpl(PRIV); + // CHECK-DAG: [[PRIV_LOAD5:%[a-zA-Z0-9]+]] = load i32*, i32* addrspace(4)* [[PRIV]].ascast + // CHECK-DAG: call spir_func void @_Z4tmplIPU3AS0iEvT_(i32* [[PRIV_LOAD5]]) + tmpl(NoAS); + // CHECK-DAG: [[NoAS_LOAD5:%[a-zA-Z0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* [[NoAS]].ascast + // CHECK-DAG: call spir_func void @_Z4tmplIPiEvT_(i32 addrspace(4)* [[NoAS_LOAD5]]) +} + +// CHECK-DAG: define linkonce_odr spir_func void @_Z4tmplIPU3AS1iEvT_(i32 addrspace(1)* % +// CHECK-DAG: define linkonce_odr spir_func void @_Z4tmplIPU3AS3iEvT_(i32 addrspace(3)* % +// CHECK-DAG: define linkonce_odr spir_func void @_Z4tmplIPU3AS0iEvT_(i32* % +// CHECK-DAG: define linkonce_odr spir_func void @_Z4tmplIPiEvT_(i32 addrspace(4)* % diff --git a/clang/test/CodeGenSYCL/address-space-deduction.cpp b/clang/test/CodeGenSYCL/address-space-deduction.cpp new file mode 100644 index 0000000..7fb5d75 --- /dev/null +++ b/clang/test/CodeGenSYCL/address-space-deduction.cpp @@ -0,0 +1,74 @@ +// RUN: %clang_cc1 -triple spir64 -fsycl-is-device -disable-llvm-passes -emit-llvm %s -o - | FileCheck %s + +// CHECK: @_ZZ4testvE3foo = internal addrspace(1) constant i32 66, align 4 +// CHECK: @[[STR:[.a-zA-Z0-9_]+]] = private unnamed_addr addrspace(1) constant [14 x i8] c"Hello, world!\00", align 1 + +// CHECK-LABEL: @_Z4testv +void test() { + static const int foo = 0x42; + + // CHECK: %i.ascast = addrspacecast i32* %i to i32 addrspace(4)* + // CHECK: %[[ARR:[a-zA-Z0-9]+]] = alloca [42 x i32] + // CHECK: %[[ARR]].ascast = addrspacecast [42 x i32]* %[[ARR]] to [42 x i32] addrspace(4)* + + int i = 0; + int *pptr = &i; + // CHECK: store i32 addrspace(4)* %i.ascast, i32 addrspace(4)* addrspace(4)* %pptr.ascast + bool is_i_ptr = (pptr == &i); + // CHECK: %[[VALPPTR:[0-9]+]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* %pptr.ascast + // CHECK: %cmp{{[0-9]*}} = icmp eq i32 addrspace(4)* %[[VALPPTR]], %i.ascast + *pptr = foo; + + int var23 = 23; + char *cp = (char *)&var23; + *cp = 41; + // CHECK: store i32 23, i32 addrspace(4)* %[[VAR:[a-zA-Z0-9.]+]] + // CHECK: [[VARCAST:%.*]] = bitcast i32 addrspace(4)* %[[VAR]] to i8 addrspace(4)* + // CHECK: store i8 addrspace(4)* [[VARCAST]], i8 addrspace(4)* addrspace(4)* %{{.*}} + + int arr[42]; + char *cpp = (char *)arr; + *cpp = 43; + // CHECK: [[ARRDECAY:%.*]] = getelementptr inbounds [42 x i32], [42 x i32] addrspace(4)* %[[ARR]].ascast, i64 0, i64 0 + // CHECK: [[ARRCAST:%.*]] = bitcast i32 addrspace(4)* [[ARRDECAY]] to i8 addrspace(4)* + // CHECK: store i8 addrspace(4)* [[ARRCAST]], i8 addrspace(4)* addrspace(4)* %{{.*}} + + int *aptr = arr + 10; + if (aptr < arr + sizeof(arr)) + *aptr = 44; + // CHECK: %[[VALAPTR:.*]] = load i32 addrspace(4)*, i32 addrspace(4)* addrspace(4)* %aptr.ascast + // CHECK: %[[ARRDCY2:.*]] = getelementptr inbounds [42 x i32], [42 x i32] addrspace(4)* %[[ARR]].ascast, i64 0, i64 0 + // CHECK: %[[ADDPTR:.*]] = getelementptr inbounds i32, i32 addrspace(4)* %[[ARRDCY2]], i64 168 + // CHECK: %cmp{{[0-9]+}} = icmp ult i32 addrspace(4)* %[[VALAPTR]], %[[ADDPTR]] + + const char *str = "Hello, world!"; + // CHECK: store i8 addrspace(4)* getelementptr inbounds ([14 x i8], [14 x i8] addrspace(4)* addrspacecast ([14 x i8] addrspace(1)* @[[STR]] to [14 x i8] addrspace(4)*), i64 0, i64 0), i8 addrspace(4)* addrspace(4)* %[[STRVAL:[a-zA-Z0-9]+]].ascast, align 8 + + i = str[0]; + + const char *phi_str = i > 2 ? str : "Another hello world!"; + (void)phi_str; + // CHECK: %[[COND:[a-zA-Z0-9]+]] = icmp sgt i32 %{{.*}}, 2 + // CHECK: br i1 %[[COND]], label %[[CONDTRUE:[.a-zA-Z0-9]+]], label %[[CONDFALSE:[.a-zA-Z0-9]+]] + + // CHECK: [[CONDTRUE]]: + // CHECK-NEXT: %[[VALTRUE:[a-zA-Z0-9]+]] = load i8 addrspace(4)*, i8 addrspace(4)* addrspace(4)* %str.ascast + // CHECK-NEXT: br label %[[CONDEND:[.a-zA-Z0-9]+]] + + // CHECK: [[CONDFALSE]]: + + // CHECK: [[CONDEND]]: + // CHECK-NEXT: phi i8 addrspace(4)* [ %[[VALTRUE]], %[[CONDTRUE]] ], [ getelementptr inbounds ([21 x i8], [21 x i8] addrspace(4)* addrspacecast ([21 x i8] addrspace(1)* @{{.*}} to [21 x i8] addrspace(4)*), i64 0, i64 0), %[[CONDFALSE]] ] + + const char *select_null = i > 2 ? "Yet another Hello world" : nullptr; + (void)select_null; + // CHECK: select i1 %{{.*}}, i8 addrspace(4)* getelementptr inbounds ([24 x i8], [24 x i8] addrspace(4)* addrspacecast ([24 x i8] addrspace(1)* @{{.*}} to [24 x i8] addrspace(4)*), i64 0, i64 0) + + const char *select_str_trivial1 = true ? str : "Another hello world!"; + (void)select_str_trivial1; + // CHECK: %[[TRIVIALTRUE:[a-zA-Z0-9]+]] = load i8 addrspace(4)*, i8 addrspace(4)* addrspace(4)* %[[STRVAL]] + // CHECK: store i8 addrspace(4)* %[[TRIVIALTRUE]], i8 addrspace(4)* addrspace(4)* %{{.*}}, align 8 + + const char *select_str_trivial2 = false ? str : "Another hello world!"; + (void)select_str_trivial2; +} diff --git a/clang/test/CodeGenSYCL/address-space-mangling.cpp b/clang/test/CodeGenSYCL/address-space-mangling.cpp new file mode 100644 index 0000000..76feec5 --- /dev/null +++ b/clang/test/CodeGenSYCL/address-space-mangling.cpp @@ -0,0 +1,30 @@ +// RUN: %clang_cc1 -triple spir64 -fsycl-is-device -disable-llvm-passes -emit-llvm %s -o - | FileCheck %s --check-prefix=SPIR +// RUN: %clang_cc1 -triple x86_64 -fsycl-is-device -disable-llvm-passes -emit-llvm %s -o - | FileCheck %s --check-prefix=X86 + +// REQUIRES: x86-registered-target + +void foo(__attribute__((opencl_global)) int *); +void foo(__attribute__((opencl_local)) int *); +void foo(__attribute__((opencl_private)) int *); +void foo(int *); + +// SPIR: declare spir_func void @_Z3fooPU3AS1i(i32 addrspace(1)*) #1 +// SPIR: declare spir_func void @_Z3fooPU3AS3i(i32 addrspace(3)*) #1 +// SPIR: declare spir_func void @_Z3fooPU3AS0i(i32*) #1 +// SPIR: declare spir_func void @_Z3fooPi(i32 addrspace(4)*) #1 + +// X86: declare void @_Z3fooPU8SYglobali(i32*) #1 +// X86: declare void @_Z3fooPU7SYlocali(i32*) #1 +// X86: declare void @_Z3fooPU9SYprivatei(i32*) #1 +// X86: declare void @_Z3fooPi(i32*) #1 + +void test() { + __attribute__((opencl_global)) int *glob; + __attribute__((opencl_local)) int *loc; + __attribute__((opencl_private)) int *priv; + int *def; + foo(glob); + foo(loc); + foo(priv); + foo(def); +} diff --git a/clang/test/SemaSYCL/address-space-conversions.cpp b/clang/test/SemaSYCL/address-space-conversions.cpp new file mode 100644 index 0000000..2183f0c --- /dev/null +++ b/clang/test/SemaSYCL/address-space-conversions.cpp @@ -0,0 +1,64 @@ +// RUN: %clang_cc1 -fsycl-is-device -verify -fsyntax-only %s + +void bar(int &Data) {} +void bar2(int &Data) {} +void bar(__attribute__((opencl_private)) int &Data) {} +void foo(int *Data) {} +void foo2(int *Data) {} +void foo(__attribute__((opencl_private)) int *Data) {} +void baz(__attribute__((opencl_private)) int *Data) {} // expected-note {{candidate function not viable: cannot pass pointer to generic address space as a pointer to address space '__private' in 1st argument}} + +template +void tmpl(T *t) {} + +void usages() { + __attribute__((opencl_global)) int *GLOB; + __attribute__((opencl_private)) int *PRIV; + __attribute__((opencl_local)) int *LOC; + int *NoAS; + + GLOB = PRIV; // expected-error {{assigning '__private int *' to '__global int *' changes address space of pointer}} + GLOB = LOC; // expected-error {{assigning '__local int *' to '__global int *' changes address space of pointer}} + PRIV = static_cast<__attribute__((opencl_private)) int *>(GLOB); // expected-error {{static_cast from '__global int *' to '__private int *' is not allowed}} + PRIV = static_cast<__attribute__((opencl_private)) int *>(LOC); // expected-error {{static_cast from '__local int *' to '__private int *' is not allowed}} + NoAS = GLOB + PRIV; // expected-error {{invalid operands to binary expression ('__global int *' and '__private int *')}} + NoAS = GLOB + LOC; // expected-error {{invalid operands to binary expression ('__global int *' and '__local int *')}} + NoAS += GLOB; // expected-error {{invalid operands to binary expression ('int *' and '__global int *')}} + + bar(*GLOB); + bar2(*GLOB); + + bar(*PRIV); + bar2(*PRIV); + + bar(*NoAS); + bar2(*NoAS); + + bar(*LOC); + bar2(*LOC); + + foo(GLOB); + foo2(GLOB); + foo(PRIV); + foo2(PRIV); + foo(NoAS); + foo2(NoAS); + foo(LOC); + foo2(LOC); + + tmpl(GLOB); + tmpl(PRIV); + tmpl(NoAS); + tmpl(LOC); + + // Implicit casts to named address space are disallowed + baz(NoAS); // expected-error {{no matching function for call to 'baz'}} + __attribute__((opencl_local)) int *l = NoAS; // expected-error {{cannot initialize a variable of type '__local int *' with an lvalue of type 'int *'}} + + (void)static_cast(GLOB); + (void)static_cast(GLOB); + int *i = GLOB; + void *v = GLOB; + (void)i; + (void)v; +} diff --git a/clang/test/SemaTemplate/address_space-dependent.cpp b/clang/test/SemaTemplate/address_space-dependent.cpp index 09ae0a6..a42e4da 100644 --- a/clang/test/SemaTemplate/address_space-dependent.cpp +++ b/clang/test/SemaTemplate/address_space-dependent.cpp @@ -43,7 +43,7 @@ void neg() { template void tooBig() { - __attribute__((address_space(I))) int *bounds; // expected-error {{address space is larger than the maximum supported (8388593)}} + __attribute__((address_space(I))) int *bounds; // expected-error {{address space is larger than the maximum supported (8388590)}} } template @@ -101,7 +101,7 @@ int main() { car<1, 2, 3>(); // expected-note {{in instantiation of function template specialization 'car<1, 2, 3>' requested here}} HasASTemplateFields<1> HASTF; neg<-1>(); // expected-note {{in instantiation of function template specialization 'neg<-1>' requested here}} - correct<0x7FFFF1>(); + correct<0x7FFFED>(); tooBig<8388650>(); // expected-note {{in instantiation of function template specialization 'tooBig<8388650>' requested here}} __attribute__((address_space(1))) char *x;