Skip to content

[clang] Use TargetInfo to determine device kernel calling convention #144728

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jun 18, 2025

Conversation

sarnex
Copy link
Member

@sarnex sarnex commented Jun 18, 2025

We should abstract this logic away to TargetInfo. See #137882 for more information.

…ntion

Signed-off-by: Sarnie, Nick <nick.sarnie@intel.com>
@sarnex sarnex marked this pull request as ready for review June 18, 2025 15:52
@llvmbot llvmbot added clang Clang issues not falling into any other category backend:AMDGPU clang:codegen IR generation bugs: mangling, exceptions, etc. labels Jun 18, 2025
@llvmbot
Copy link
Member

llvmbot commented Jun 18, 2025

@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-backend-amdgpu

Author: Nick Sarnie (sarnex)

Changes

We should abstract this logic away to TargetInfo. See #137882 for more information.


Full diff: https://github.com/llvm/llvm-project/pull/144728.diff

6 Files Affected:

  • (modified) clang/lib/CodeGen/CGCall.cpp (+2-11)
  • (modified) clang/lib/CodeGen/TargetInfo.cpp (+4-4)
  • (modified) clang/lib/CodeGen/TargetInfo.h (+2-2)
  • (modified) clang/lib/CodeGen/Targets/AMDGPU.cpp (+2-2)
  • (modified) clang/lib/CodeGen/Targets/NVPTX.cpp (+1-1)
  • (modified) clang/lib/CodeGen/Targets/SPIR.cpp (+2-2)
diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index a06455d25b1ef..fd75de42515da 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -83,17 +83,8 @@ unsigned CodeGenTypes::ClangCallConvToLLVMCallConv(CallingConv CC) {
     return llvm::CallingConv::AArch64_SVE_VectorCall;
   case CC_SpirFunction:
     return llvm::CallingConv::SPIR_FUNC;
-  case CC_DeviceKernel: {
-    if (CGM.getLangOpts().OpenCL)
-      return CGM.getTargetCodeGenInfo().getOpenCLKernelCallingConv();
-    if (CGM.getTriple().isSPIROrSPIRV())
-      return llvm::CallingConv::SPIR_KERNEL;
-    if (CGM.getTriple().isAMDGPU())
-      return llvm::CallingConv::AMDGPU_KERNEL;
-    if (CGM.getTriple().isNVPTX())
-      return llvm::CallingConv::PTX_Kernel;
-    llvm_unreachable("Unknown kernel calling convention");
-  }
+  case CC_DeviceKernel:
+    return CGM.getTargetCodeGenInfo().getDeviceKernelCallingConv();
   case CC_PreserveMost:
     return llvm::CallingConv::PreserveMost;
   case CC_PreserveAll:
diff --git a/clang/lib/CodeGen/TargetInfo.cpp b/clang/lib/CodeGen/TargetInfo.cpp
index f3df92c44bb6b..bc771c45951b3 100644
--- a/clang/lib/CodeGen/TargetInfo.cpp
+++ b/clang/lib/CodeGen/TargetInfo.cpp
@@ -103,11 +103,11 @@ TargetCodeGenInfo::getDependentLibraryOption(llvm::StringRef Lib,
   Opt += Lib;
 }
 
-unsigned TargetCodeGenInfo::getOpenCLKernelCallingConv() const {
-  // OpenCL kernels are called via an explicit runtime API with arguments
-  // set with clSetKernelArg(), not as normal sub-functions.
+unsigned TargetCodeGenInfo::getDeviceKernelCallingConv() const {
+  // Device kernels are called via an explicit runtime API with arguments,
+  // such as set with clSetKernelArg() for OpenCL, not as normal sub-functions.
   // Return SPIR_KERNEL by default as the kernel calling convention to
-  // ensure the fingerprint is fixed such way that each OpenCL argument
+  // ensure the fingerprint is fixed such way that each kernel argument
   // gets one matching argument in the produced kernel function argument
   // list to enable feasible implementation of clSetKernelArg() with
   // aggregates etc. In case we would use the default C calling conv here,
diff --git a/clang/lib/CodeGen/TargetInfo.h b/clang/lib/CodeGen/TargetInfo.h
index 2783e222eb802..b4057d369f988 100644
--- a/clang/lib/CodeGen/TargetInfo.h
+++ b/clang/lib/CodeGen/TargetInfo.h
@@ -298,8 +298,8 @@ class TargetCodeGenInfo {
                                        llvm::StringRef Value,
                                        llvm::SmallString<32> &Opt) const {}
 
-  /// Get LLVM calling convention for OpenCL kernel.
-  virtual unsigned getOpenCLKernelCallingConv() const;
+  /// Get LLVM calling convention for device kernels.
+  virtual unsigned getDeviceKernelCallingConv() const;
 
   /// Get target specific null pointer.
   /// \param T is the LLVM type of the null pointer.
diff --git a/clang/lib/CodeGen/Targets/AMDGPU.cpp b/clang/lib/CodeGen/Targets/AMDGPU.cpp
index 8660373c3927f..47a552a7bf495 100644
--- a/clang/lib/CodeGen/Targets/AMDGPU.cpp
+++ b/clang/lib/CodeGen/Targets/AMDGPU.cpp
@@ -304,7 +304,7 @@ class AMDGPUTargetCodeGenInfo : public TargetCodeGenInfo {
 
   void setTargetAttributes(const Decl *D, llvm::GlobalValue *GV,
                            CodeGen::CodeGenModule &M) const override;
-  unsigned getOpenCLKernelCallingConv() const override;
+  unsigned getDeviceKernelCallingConv() const override;
 
   llvm::Constant *getNullPointer(const CodeGen::CodeGenModule &CGM,
       llvm::PointerType *T, QualType QT) const override;
@@ -431,7 +431,7 @@ void AMDGPUTargetCodeGenInfo::setTargetAttributes(
     F->addFnAttr("amdgpu-ieee", "false");
 }
 
-unsigned AMDGPUTargetCodeGenInfo::getOpenCLKernelCallingConv() const {
+unsigned AMDGPUTargetCodeGenInfo::getDeviceKernelCallingConv() const {
   return llvm::CallingConv::AMDGPU_KERNEL;
 }
 
diff --git a/clang/lib/CodeGen/Targets/NVPTX.cpp b/clang/lib/CodeGen/Targets/NVPTX.cpp
index ad802c9131de0..82bdfe2666b52 100644
--- a/clang/lib/CodeGen/Targets/NVPTX.cpp
+++ b/clang/lib/CodeGen/Targets/NVPTX.cpp
@@ -78,7 +78,7 @@ class NVPTXTargetCodeGenInfo : public TargetCodeGenInfo {
     return true;
   }
 
-  unsigned getOpenCLKernelCallingConv() const override {
+  unsigned getDeviceKernelCallingConv() const override {
     return llvm::CallingConv::PTX_Kernel;
   }
 
diff --git a/clang/lib/CodeGen/Targets/SPIR.cpp b/clang/lib/CodeGen/Targets/SPIR.cpp
index 2f1e43cdc8cc3..afa23bffcd073 100644
--- a/clang/lib/CodeGen/Targets/SPIR.cpp
+++ b/clang/lib/CodeGen/Targets/SPIR.cpp
@@ -51,7 +51,7 @@ class CommonSPIRTargetCodeGenInfo : public TargetCodeGenInfo {
         getABIInfo().getDataLayout().getAllocaAddrSpace());
   }
 
-  unsigned getOpenCLKernelCallingConv() const override;
+  unsigned getDeviceKernelCallingConv() const override;
   llvm::Type *getOpenCLType(CodeGenModule &CGM, const Type *T) const override;
   llvm::Type *
   getHLSLType(CodeGenModule &CGM, const Type *Ty,
@@ -219,7 +219,7 @@ void computeSPIRKernelABIInfo(CodeGenModule &CGM, CGFunctionInfo &FI) {
 }
 }
 
-unsigned CommonSPIRTargetCodeGenInfo::getOpenCLKernelCallingConv() const {
+unsigned CommonSPIRTargetCodeGenInfo::getDeviceKernelCallingConv() const {
   return llvm::CallingConv::SPIR_KERNEL;
 }
 

@sarnex sarnex changed the title [clang][NFCI] Use TargetInfo to determine device kernel calling convention [clang] Use TargetInfo to determine device kernel calling convention Jun 18, 2025
Signed-off-by: Sarnie, Nick <nick.sarnie@intel.com>
@sarnex sarnex merged commit 86d1d6b into llvm:main Jun 18, 2025
7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
backend:AMDGPU clang:codegen IR generation bugs: mangling, exceptions, etc. clang Clang issues not falling into any other category
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants