[clang] Simplify device kernel attributes #137882

sarnex · 2025-04-29T21:45:14Z

We have multiple different attributes in clang representing device kernels for specific targets/languages. Refactor them into one attribute with different spellings to make it more easily scalable for new languages/targets.

github-actions · 2025-04-30T17:29:19Z

✅ With the latest revision this PR passed the C/C++ code formatter.

llvmbot · 2025-05-01T14:15:01Z

@llvm/pr-subscribers-backend-amdgpu
@llvm/pr-subscribers-backend-aarch64
@llvm/pr-subscribers-llvm-binary-utilities
@llvm/pr-subscribers-debuginfo
@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-clang

Author: Nick Sarnie (sarnex)

Changes

SPIR kernels have a specific calling convention, so add an attribute to explicitly specify this calling convention when targetting pure SPIR/SPIRV.

Full diff: https://github.com/llvm/llvm-project/pull/137882.diff

14 Files Affected:

(modified) clang/include/clang/Basic/Attr.td (+8)
(modified) clang/include/clang/Basic/Specifiers.h (+1)
(modified) clang/lib/AST/ItaniumMangle.cpp (+1)
(modified) clang/lib/AST/Type.cpp (+2)
(modified) clang/lib/AST/TypePrinter.cpp (+3)
(modified) clang/lib/Basic/Targets/SPIR.h (+4-2)
(modified) clang/lib/CodeGen/CGCall.cpp (+2)
(modified) clang/lib/CodeGen/CGDebugInfo.cpp (+2)
(modified) clang/lib/CodeGen/Targets/SPIR.cpp (+17)
(modified) clang/lib/Sema/SemaDeclAttr.cpp (+6)
(modified) clang/test/Misc/pragma-attribute-supported-attributes-list.test (+1)
(added) clang/test/Misc/spir-kernel-attr.c (+10)
(modified) llvm/include/llvm/BinaryFormat/Dwarf.def (+1)
(modified) llvm/include/llvm/DebugInfo/DWARF/DWARFTypePrinter.h (+3)

diff --git a/clang/include/clang/Basic/Attr.td b/clang/include/clang/Basic/Attr.td
index a734eb6658c3d..c3a3db095bb81 100644
--- a/clang/include/clang/Basic/Attr.td
+++ b/clang/include/clang/Basic/Attr.td
@@ -472,7 +472,9 @@ def TargetRISCV : TargetArch<["riscv32", "riscv64"]>;
 def TargetX86 : TargetArch<["x86"]>;
 def TargetX86_64 : TargetArch<["x86_64"]>;
 def TargetAnyX86 : TargetArch<["x86", "x86_64"]>;
+def TargetSPIR : TargetArch<["spir", "spir64"]>;
 def TargetSPIRV : TargetArch<["spirv", "spirv32", "spirv64"]>;
+def TargetAnySPIR : TargetArch<!listconcat(TargetSPIR.Arches, TargetSPIRV.Arches)>;
 def TargetWebAssembly : TargetArch<["wasm32", "wasm64"]>;
 def TargetNVPTX : TargetArch<["nvptx", "nvptx64"]>;
 def TargetWindows : TargetSpec {
@@ -1504,6 +1506,12 @@ def NVPTXKernel : InheritableAttr, TargetSpecificAttr<TargetNVPTX> {
   let Documentation = [Undocumented];
 }
 
+def SPIRKernel : InheritableAttr, TargetSpecificAttr<TargetAnySPIR> {
+  let Spellings = [Clang<"spir_kernel">];
+  let Subjects = SubjectList<[Function]>;
+  let Documentation = [Undocumented];
+}
+
 def HIPManaged : InheritableAttr {
   let Spellings = [GNU<"managed">, Declspec<"__managed__">];
   let Subjects = SubjectList<[Var]>;
diff --git a/clang/include/clang/Basic/Specifiers.h b/clang/include/clang/Basic/Specifiers.h
index 491badcc804e7..f9a72f378490e 100644
--- a/clang/include/clang/Basic/Specifiers.h
+++ b/clang/include/clang/Basic/Specifiers.h
@@ -289,6 +289,7 @@ namespace clang {
     CC_AAPCS_VFP,          // __attribute__((pcs("aapcs-vfp")))
     CC_IntelOclBicc,       // __attribute__((intel_ocl_bicc))
     CC_SpirFunction,       // default for OpenCL functions on SPIR target
+    CC_SpirKernel,         // __attribute__((spir_kernel))
     CC_OpenCLKernel,       // inferred for OpenCL kernels
     CC_Swift,              // __attribute__((swiftcall))
     CC_SwiftAsync,         // __attribute__((swiftasynccall))
diff --git a/clang/lib/AST/ItaniumMangle.cpp b/clang/lib/AST/ItaniumMangle.cpp
index 33a8728728574..c16005ccaf51f 100644
--- a/clang/lib/AST/ItaniumMangle.cpp
+++ b/clang/lib/AST/ItaniumMangle.cpp
@@ -3532,6 +3532,7 @@ StringRef CXXNameMangler::getCallingConvQualifierName(CallingConv CC) {
   case CC_AMDGPUKernelCall:
   case CC_IntelOclBicc:
   case CC_SpirFunction:
+  case CC_SpirKernel:
   case CC_OpenCLKernel:
   case CC_PreserveMost:
   case CC_PreserveAll:
diff --git a/clang/lib/AST/Type.cpp b/clang/lib/AST/Type.cpp
index 59369fba2e772..da9ba8e3e971a 100644
--- a/clang/lib/AST/Type.cpp
+++ b/clang/lib/AST/Type.cpp
@@ -3623,6 +3623,8 @@ StringRef FunctionType::getNameForCallConv(CallingConv CC) {
     return "intel_ocl_bicc";
   case CC_SpirFunction:
     return "spir_function";
+  case CC_SpirKernel:
+    return "spir_kernel";
   case CC_OpenCLKernel:
     return "opencl_kernel";
   case CC_Swift:
diff --git a/clang/lib/AST/TypePrinter.cpp b/clang/lib/AST/TypePrinter.cpp
index cba1a2d98d660..5aa37ca9ec17f 100644
--- a/clang/lib/AST/TypePrinter.cpp
+++ b/clang/lib/AST/TypePrinter.cpp
@@ -1115,6 +1115,9 @@ void TypePrinter::printFunctionAfter(const FunctionType::ExtInfo &Info,
     case CC_OpenCLKernel:
       // Do nothing. These CCs are not available as attributes.
       break;
+    case CC_SpirKernel:
+      OS << " __attribute__((spir_kernel))";
+      break;
     case CC_Swift:
       OS << " __attribute__((swiftcall))";
       break;
diff --git a/clang/lib/Basic/Targets/SPIR.h b/clang/lib/Basic/Targets/SPIR.h
index bf249e271a870..43c71a1aeb5d5 100644
--- a/clang/lib/Basic/Targets/SPIR.h
+++ b/clang/lib/Basic/Targets/SPIR.h
@@ -191,8 +191,10 @@ class LLVM_LIBRARY_VISIBILITY BaseSPIRTargetInfo : public TargetInfo {
   }
 
   CallingConvCheckResult checkCallingConvention(CallingConv CC) const override {
-    return (CC == CC_SpirFunction || CC == CC_OpenCLKernel) ? CCCR_OK
-                                                            : CCCR_Warning;
+    return (CC == CC_SpirKernel || CC == CC_SpirFunction ||
+            CC == CC_OpenCLKernel)
+               ? CCCR_OK
+               : CCCR_Warning;
   }
 
   CallingConv getDefaultCallingConv() const override {
diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 83b0e8e965770..a7bbff116b584 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -84,6 +84,8 @@ unsigned CodeGenTypes::ClangCallConvToLLVMCallConv(CallingConv CC) {
     return llvm::CallingConv::AMDGPU_KERNEL;
   case CC_SpirFunction:
     return llvm::CallingConv::SPIR_FUNC;
+  case CC_SpirKernel:
+    return llvm::CallingConv::SPIR_KERNEL;
   case CC_OpenCLKernel:
     return CGM.getTargetCodeGenInfo().getOpenCLKernelCallingConv();
   case CC_PreserveMost:
diff --git a/clang/lib/CodeGen/CGDebugInfo.cpp b/clang/lib/CodeGen/CGDebugInfo.cpp
index f3ec498d4064b..10fba7b772486 100644
--- a/clang/lib/CodeGen/CGDebugInfo.cpp
+++ b/clang/lib/CodeGen/CGDebugInfo.cpp
@@ -1589,6 +1589,8 @@ static unsigned getDwarfCC(CallingConv CC) {
     return llvm::dwarf::DW_CC_LLVM_IntelOclBicc;
   case CC_SpirFunction:
     return llvm::dwarf::DW_CC_LLVM_SpirFunction;
+  case CC_SpirKernel:
+    return llvm::dwarf::DW_CC_LLVM_SpirKernel;
   case CC_OpenCLKernel:
   case CC_AMDGPUKernelCall:
     return llvm::dwarf::DW_CC_LLVM_OpenCLKernel;
diff --git a/clang/lib/CodeGen/Targets/SPIR.cpp b/clang/lib/CodeGen/Targets/SPIR.cpp
index f35c124f50aa0..29a258e457317 100644
--- a/clang/lib/CodeGen/Targets/SPIR.cpp
+++ b/clang/lib/CodeGen/Targets/SPIR.cpp
@@ -60,6 +60,8 @@ class CommonSPIRTargetCodeGenInfo : public TargetCodeGenInfo {
       llvm::Type *ElementType, llvm::LLVMContext &Ctx) const;
   void
   setOCLKernelStubCallingConvention(const FunctionType *&FT) const override;
+  void setTargetAttributes(const Decl *D, llvm::GlobalValue *GV,
+                           CodeGen::CodeGenModule &M) const override;
 };
 class SPIRVTargetCodeGenInfo : public CommonSPIRTargetCodeGenInfo {
 public:
@@ -238,6 +240,20 @@ void CommonSPIRTargetCodeGenInfo::setOCLKernelStubCallingConvention(
       FT, FT->getExtInfo().withCallingConv(CC_SpirFunction));
 }
 
+void CommonSPIRTargetCodeGenInfo::setTargetAttributes(
+    const Decl *D, llvm::GlobalValue *GV, CodeGen::CodeGenModule &M) const {
+  const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(D);
+  if (!FD)
+    return;
+
+  llvm::Function *F = cast<llvm::Function>(GV);
+
+  // Attach kernel metadata directly if compiling for SPIR.
+  if (FD->hasAttr<SPIRKernelAttr>()) {
+    F->setCallingConv(llvm::CallingConv::SPIR_KERNEL);
+  }
+}
+
 LangAS
 SPIRVTargetCodeGenInfo::getGlobalVarAddressSpace(CodeGenModule &CGM,
                                                  const VarDecl *D) const {
@@ -262,6 +278,7 @@ SPIRVTargetCodeGenInfo::getGlobalVarAddressSpace(CodeGenModule &CGM,
 
 void SPIRVTargetCodeGenInfo::setTargetAttributes(
     const Decl *D, llvm::GlobalValue *GV, CodeGen::CodeGenModule &M) const {
+  CommonSPIRTargetCodeGenInfo::setTargetAttributes(D, GV, M);
   if (!M.getLangOpts().HIP ||
       M.getTarget().getTriple().getVendor() != llvm::Triple::AMD)
     return;
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index ab66ae860f86b..56aea1e878f23 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -5503,6 +5503,9 @@ bool Sema::CheckCallingConvAttr(const ParsedAttr &Attrs, CallingConv &CC,
                                   llvm::Log2_64(ABIVLen) - 5);
     break;
   }
+  case ParsedAttr::AT_SPIRKernel:
+    CC = CC_SpirKernel;
+    break;
   default: llvm_unreachable("unexpected attribute kind");
   }
 
@@ -7152,6 +7155,9 @@ ProcessDeclAttribute(Sema &S, Scope *scope, Decl *D, const ParsedAttr &AL,
   case ParsedAttr::AT_CUDALaunchBounds:
     handleLaunchBoundsAttr(S, D, AL);
     break;
+  case ParsedAttr::AT_SPIRKernel:
+    handleSimpleAttribute<SPIRKernelAttr>(S, D, AL);
+    break;
   case ParsedAttr::AT_Restrict:
     handleRestrictAttr(S, D, AL);
     break;
diff --git a/clang/test/Misc/pragma-attribute-supported-attributes-list.test b/clang/test/Misc/pragma-attribute-supported-attributes-list.test
index 7affacb1a109a..53a7ea4c8033b 100644
--- a/clang/test/Misc/pragma-attribute-supported-attributes-list.test
+++ b/clang/test/Misc/pragma-attribute-supported-attributes-list.test
@@ -181,6 +181,7 @@
 // CHECK-NEXT: ReturnTypestate (SubjectMatchRule_function, SubjectMatchRule_variable_is_parameter)
 // CHECK-NEXT: ReturnsNonNull (SubjectMatchRule_objc_method, SubjectMatchRule_function)
 // CHECK-NEXT: ReturnsTwice (SubjectMatchRule_function)
+// CHECK-NEXT: SPIRKernel (SubjectMatchRule_function)
 // CHECK-NEXT: SYCLKernelEntryPoint (SubjectMatchRule_function)
 // CHECK-NEXT: SYCLSpecialClass (SubjectMatchRule_record)
 // CHECK-NEXT: ScopedLockable (SubjectMatchRule_record)
diff --git a/clang/test/Misc/spir-kernel-attr.c b/clang/test/Misc/spir-kernel-attr.c
new file mode 100644
index 0000000000000..40de980716ff2
--- /dev/null
+++ b/clang/test/Misc/spir-kernel-attr.c
@@ -0,0 +1,10 @@
+// RUN: %clang_cc1 -triple spir -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 -triple spir64 -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 -triple spirv-unknown-vulkan-compute -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 -triple spirv32 -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 -triple spirv64 -emit-llvm %s -o - | FileCheck %s
+__attribute__((spir_kernel)) void foo(void) {}
+
+[[clang::spir_kernel]] void bar(void) {}
+
+// CHECK-COUNT-2: spir_kernel
diff --git a/llvm/include/llvm/BinaryFormat/Dwarf.def b/llvm/include/llvm/BinaryFormat/Dwarf.def
index e52324a8ebc12..575ab05f4e3c4 100644
--- a/llvm/include/llvm/BinaryFormat/Dwarf.def
+++ b/llvm/include/llvm/BinaryFormat/Dwarf.def
@@ -1127,6 +1127,7 @@ HANDLE_DW_CC(0xcd, LLVM_PreserveNone)
 HANDLE_DW_CC(0xce, LLVM_RISCVVectorCall)
 HANDLE_DW_CC(0xcf, LLVM_SwiftTail)
 HANDLE_DW_CC(0xd0, LLVM_RISCVVLSCall)
+HANDLE_DW_CC(0xd1, LLVM_SpirKernel)
 // From GCC source code (include/dwarf2.h): This DW_CC_ value is not currently
 // generated by any toolchain.  It is used internally to GDB to indicate OpenCL
 // C functions that have been compiled with the IBM XL C for OpenCL compiler and
diff --git a/llvm/include/llvm/DebugInfo/DWARF/DWARFTypePrinter.h b/llvm/include/llvm/DebugInfo/DWARF/DWARFTypePrinter.h
index bd25f6c30ebf1..82337774f2396 100644
--- a/llvm/include/llvm/DebugInfo/DWARF/DWARFTypePrinter.h
+++ b/llvm/include/llvm/DebugInfo/DWARF/DWARFTypePrinter.h
@@ -741,6 +741,9 @@ void DWARFTypePrinter<DieType>::appendSubroutineNameAfter(
       // instantiated with function types with these calling conventions won't
       // have distinct names - so we'd need to fix that too)
       break;
+    case dwarf::CallingConvention::DW_CC_LLVM_SpirKernel:
+      OS << " __attribute((spir_kernel))";
+      break;
     case dwarf::CallingConvention::DW_CC_LLVM_Swift:
       // SwiftAsync missing
       OS << " __attribute__((swiftcall))";

erichkeane · 2025-05-01T14:37:05Z

clang/lib/Sema/SemaDeclAttr.cpp

@@ -5503,6 +5503,9 @@ bool Sema::CheckCallingConvAttr(const ParsedAttr &Attrs, CallingConv &CC,
                                  llvm::Log2_64(ABIVLen) - 5);
    break;
  }
+  case ParsedAttr::AT_SPIRKernel:


Perhaps a silly request, but please sort this as 'not last'. It is too easy to miss when after a big block.

erichkeane · 2025-05-01T14:38:39Z

clang/include/clang/Basic/Attr.td

+def SPIRKernel : InheritableAttr, TargetSpecificAttr<TargetAnySPIR> {
+  let Spellings = [Clang<"spir_kernel">];
+  let Subjects = SubjectList<[Function]>;
+  let Documentation = [Undocumented];


Why is this non-documented? WE very much don't approve/accept new undocumented attributes without good reason.

Thanks, I wasn't sure how strict we were with that, will add documentation.

AaronBallman

Do we need yet another calling convention or is there a way we can start to combine some of these for all the offloading languages? It seems we keep re-implementing the same concepts multiple times for each language and it would be nice to share as much of the frontend logic as possible rather than keep doing this dance of adding another case to every switch in the compiler. :-D

sarnex · 2025-05-01T15:06:12Z

Do we need yet another calling convention or is there a way we can start to combine some of these for all the offloading languages? It seems we keep re-implementing the same concepts multiple times for each language and it would be nice to share as much of the frontend logic as possible rather than keep doing this dance of adding another case to every switch in the compiler. :-D

Do you have an idea? Maybe we could have a calling_convention attribute that takes in a string that specifies the calling convention?

AaronBallman · 2025-05-01T15:21:26Z

Do we need yet another calling convention or is there a way we can start to combine some of these for all the offloading languages? It seems we keep re-implementing the same concepts multiple times for each language and it would be nice to share as much of the frontend logic as possible rather than keep doing this dance of adding another case to every switch in the compiler. :-D

Do you have an idea? Maybe we could have a calling_convention attribute that takes in a string that specifies the calling convention?

I was thinking that orthogonal calling conventions could be combined so that we need fewer of them. e.g., if spir_kernel cannot be mixed with nvptx_kernel in the same TU, then there's no need for two distinct internal representations for the attributes, just two distinct spellings that are based on the target.

sarnex · 2025-05-01T15:23:43Z

I was thinking that orthogonal calling conventions could be combined so that we need fewer of them. e.g., if spir_kernel cannot be mixed with nvptx_kernel in the same TU, then there's no need for two distinct internal representations for the attributes, just two distinct spellings that are based on the target.

Ah, I see.

Okay, I'll try, but it might take me a while. Thanks for the feedback. Marking this as a draft for now.

Signed-off-by: Sarnie, Nick <[email protected]>

sarnex · 2025-05-07T21:02:55Z

@AaronBallman I just pushed a first attempt to combine the attrs, however I'm not sure if it's much cleaner given the attrs have different subjects/allowed cases/expected warnings/etc. Do you mind taking a first look and seeing if you think the general direction is a good idea? Thanks

AaronBallman

CC @erichkeane and @jdoerfert for other opinions on the direction this is heading, but my initial thoughts are that this is not unreasonable. But I'd love to hear from others!

AaronBallman · 2025-05-08T12:13:40Z

clang/include/clang/Basic/Attr.td

+   inline bool isAMDGPUSpelling() const {
+     return isAMDGPUSpelling(*this);
+   }
+   template<typename T>
+   static inline bool isAMDGPUSpelling(const T& Attr) {
+     return Attr.getAttrName()->getName() == "amdgpu_kernel";
+   }
+   inline bool isNVPTXSpelling() const {
+     return isNVPTXSpelling(*this);
+   }


There's an easier way to do this, I think. We have:

let Accessors = [Accessor<"isAMDGPU", [Clang<"amdgpu_kernel">]>, Accessor<"isNVPTX", [Clang<"nvptx_kernel">]>];

which generates an accessor method for you which returns true/false based on the spelling.

Ah cool, thanks! Will do! I think I still need some of the methods to deal with the fact ParsedAttr isn't a subclass of Attr but also has a getAttrName() function, but maybe we can get rid of half of them

I tried this, but we actually need to check the spelling mostly for ParsedAttr, so the accessors here won't help. I implemented Erich's idea to use the enum instead of the string. Let me know if you have some tablegen magic that we can use!

AaronBallman · 2025-05-08T12:15:09Z

clang/include/clang/Basic/Attr.td

+  let Spellings = [Clang<"device_kernel">, Clang<"sycl_kernel">,
+                   Clang<"nvptx_kernel">, Clang<"amdgpu_kernel">,
+                   CustomKeyword<"__kernel">, CustomKeyword<"kernel">];
+  let LangOpts = [];


Can drop this entirely.

erichkeane

I honestly really like the approach. Those attributes really only differ by spelling and are almost semantically identical, so it makes a lot of sense to do it this way. So I'm happy.

The accessors could be cleaned up, otherwise this is a LGTM.

erichkeane · 2025-05-08T12:59:10Z

clang/include/clang/Basic/Attr.td

+   }
+   template<typename T>
+   static inline bool isOpenCLSpelling(const T& Attr) {
+   return Attr.getAttrName()->getName() == "kernel" ||


Even without the Accessors thing that Aaron suggests, you can check the spelling-id vs the generated enums which is better than string compares.

Will try if it I hit an issue with Aaron's suggestion, thanks!

I ended up doing it this way using the enums.

sarnex · 2025-05-08T17:42:25Z

Thanks a lot guys, I didn't expect this to be a good direction :)

Signed-off-by: Sarnie, Nick <[email protected]>

arsenm · 2025-05-08T20:23:34Z

clang/lib/CodeGen/Targets/SPIR.cpp

@@ -227,7 +227,7 @@ void SPIRVTargetCodeGenInfo::setCUDAKernelCallingConvention(
  // Convert HIP kernels to SPIR-V kernels.
  if (getABIInfo().getContext().getLangOpts().HIP) {
    FT = getABIInfo().getContext().adjustFunctionType(
-        FT, FT->getExtInfo().withCallingConv(CC_OpenCLKernel));
+        FT, FT->getExtInfo().withCallingConv(CC_DeviceKernel));


Not sure setCUDAKernelCallingConvention still needs to exist with this?

arsenm · 2025-05-08T20:24:25Z

clang/lib/CodeGen/Targets/NVPTX.cpp

-  if (FD->hasAttr<NVPTXKernelAttr>()) {
+  if (FD->hasAttr<DeviceKernelAttr>()) {
    F->setCallingConv(llvm::CallingConv::PTX_Kernel);


This is now set in the 2 language paths,, there's additional code that can be folded now

clang/include/clang/Basic/Attr.td

AaronBallman · 2025-05-09T12:36:29Z

clang/include/clang/Basic/Attr.td

+def DeviceKernel : DeclOrTypeAttr {
+  let Spellings = [Clang<"device_kernel">, Clang<"sycl_kernel">,
+                   Clang<"nvptx_kernel">, Clang<"amdgpu_kernel">,
+                   CustomKeyword<"__kernel">, CustomKeyword<"kernel">];
  let Documentation = [SYCLKernelDocs];


I suspect we'll need to update the documentation because it's now going to list all of the various kernel attributes under SYCL kernel docs.

Good catch, will do, thanks!

sarnex closed this Apr 29, 2025

sarnex reopened this Apr 29, 2025

sarnex force-pushed the spirattr branch 3 times, most recently from 711a62e to 20498a3 Compare April 30, 2025 17:26

sarnex force-pushed the spirattr branch from 20498a3 to be91732 Compare April 30, 2025 21:00

sarnex marked this pull request as ready for review May 1, 2025 14:14

llvmbot added clang Clang issues not falling into any other category clang:frontend Language frontend issues, e.g. anything involving "Sema" clang:codegen IR generation bugs: mangling, exceptions, etc. debuginfo llvm:binary-utilities labels May 1, 2025

sarnex requested review from AaronBallman, erichkeane, AlexVlx and VyacheslavLevytskyy May 1, 2025 14:15

erichkeane reviewed May 1, 2025

View reviewed changes

AaronBallman reviewed May 1, 2025

View reviewed changes

sarnex marked this pull request as draft May 1, 2025 15:23

[clang] Simplify device kernel attributes

080a9d4

Signed-off-by: Sarnie, Nick <[email protected]>

sarnex force-pushed the spirattr branch from be91732 to 080a9d4 Compare May 7, 2025 21:01

AaronBallman reviewed May 8, 2025

View reviewed changes

erichkeane reviewed May 8, 2025

View reviewed changes

AlexVlx requested review from arsenm and yxsamliu May 8, 2025 17:49

sarnex added 3 commits May 8, 2025 12:53

apply feedback and rework member functions

0c75247

Signed-off-by: Sarnie, Nick <[email protected]>

Merge remote-tracking branch 'upstream/main'

1036640

fix merge

47ae76a

Signed-off-by: Sarnie, Nick <[email protected]>

sarnex changed the title ~~[clang] Add spir_kernel attribute~~ [clang] Simplify device kernel attributes May 8, 2025

sarnex marked this pull request as ready for review May 8, 2025 20:14

llvmbot added backend:ARM backend:AArch64 backend:AMDGPU backend:SystemZ clang:as-a-library libclang and C++ API labels May 8, 2025

arsenm reviewed May 8, 2025

View reviewed changes

AaronBallman reviewed May 9, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[clang] Simplify device kernel attributes #137882

[clang] Simplify device kernel attributes #137882

sarnex commented Apr 29, 2025 •

edited

Loading

github-actions bot commented Apr 30, 2025 •

edited

Loading

llvmbot commented May 1, 2025 •

edited

Loading

erichkeane May 1, 2025

erichkeane May 1, 2025

sarnex May 1, 2025

AaronBallman left a comment

sarnex commented May 1, 2025

AaronBallman commented May 1, 2025

sarnex commented May 1, 2025

sarnex commented May 7, 2025

AaronBallman left a comment

AaronBallman May 8, 2025

sarnex May 8, 2025 •

edited

Loading

sarnex May 8, 2025

AaronBallman May 8, 2025

erichkeane left a comment

erichkeane May 8, 2025

sarnex May 8, 2025

sarnex May 8, 2025

sarnex commented May 8, 2025

arsenm May 8, 2025

arsenm May 8, 2025

AaronBallman May 9, 2025

sarnex May 9, 2025

[clang] Simplify device kernel attributes #137882

Are you sure you want to change the base?

[clang] Simplify device kernel attributes #137882

Conversation

sarnex commented Apr 29, 2025 • edited Loading

github-actions bot commented Apr 30, 2025 • edited Loading

llvmbot commented May 1, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AaronBallman left a comment

Choose a reason for hiding this comment

sarnex commented May 1, 2025

AaronBallman commented May 1, 2025

sarnex commented May 1, 2025

sarnex commented May 7, 2025

AaronBallman left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sarnex May 8, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

erichkeane left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sarnex commented May 8, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sarnex commented Apr 29, 2025 •

edited

Loading

github-actions bot commented Apr 30, 2025 •

edited

Loading

llvmbot commented May 1, 2025 •

edited

Loading

sarnex May 8, 2025 •

edited

Loading