AMDGPU/SI: Use amdgpu_kernel calling convention for OpenCL kernels.

[• tstellarAMD]

LGTM.

cfe/trunk/lib/CodeGen/TargetInfo.cpp

 void AMDGPUTargetCodeGenInfo::setTargetAttributes(
   const Decl *D,
   llvm::GlobalValue *GV,
   CodeGen::CodeGenModule &M) const {
   const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(D);
   if (!FD)
     return;
 
+  if (M.getLangOpts().OpenCL) {
+    if (FD->hasAttr<OpenCLKernelAttr>()) {
+      // Set amdgpu_kernel calling convention for OpenCL kernels.
+      llvm::Function *Fn = cast<llvm::Function>(GV);
+      Fn->setCallingConv(llvm::CallingConv::AMDGPU_KERNEL);
+    }
+  }
+
   if (const auto Attr = FD->getAttr<AMDGPUNumVGPRAttr>()) {
     llvm::Function *F = cast<llvm::Function>(GV);
     uint32_t NumVGPR = Attr->getNumVGPR();
     if (NumVGPR != 0)
       F->addFnAttr("amdgpu_num_vgpr", llvm::utostr(NumVGPR));
   }
 
   if (const auto Attr = FD->getAttr<AMDGPUNumSGPRAttr>()) {

cfe/trunk/test/CodeGenOpenCL/amdgpu-calling-conv.cl

+// REQUIRES: amdgpu-registered-target
+// RUN: %clang_cc1 -triple amdgcn-unknown-unknown -S -emit-llvm -o - %s | FileCheck %s
+
+// CHECK: define amdgpu_kernel void @calling_conv_amdgpu_kernel()
+kernel void calling_conv_amdgpu_kernel()
+{
+}
+
+// CHECK: define void @calling_conv_none()
+void calling_conv_none()
+{
+}

cfe/trunk/test/CodeGenOpenCL/amdgpu-num-gpr-attr.cl

 // RUN: %clang_cc1 -triple amdgcn-- -target-cpu tahiti -O0 -emit-llvm -o - %s | FileCheck %s
 // RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -O0 -emit-llvm -verify -o - %s | FileCheck -check-prefix=X86 %s
 
 // Make sure this is silently accepted on other targets.
 
 __attribute__((amdgpu_num_vgpr(64))) // expected-no-diagnostics
 kernel void test_num_vgpr64() {
-// CHECK: define void @test_num_vgpr64() [[ATTR_VGPR64:#[0-9]+]]
+// CHECK: define amdgpu_kernel void @test_num_vgpr64() [[ATTR_VGPR64:#[0-9]+]]
 }
 
 __attribute__((amdgpu_num_sgpr(32))) // expected-no-diagnostics
 kernel void test_num_sgpr32() {
-// CHECK: define void @test_num_sgpr32() [[ATTR_SGPR32:#[0-9]+]]
+// CHECK: define amdgpu_kernel void @test_num_sgpr32() [[ATTR_SGPR32:#[0-9]+]]
 }
 
 __attribute__((amdgpu_num_vgpr(64), amdgpu_num_sgpr(32))) // expected-no-diagnostics
 kernel void test_num_vgpr64_sgpr32() {
-// CHECK: define void @test_num_vgpr64_sgpr32() [[ATTR_VGPR64_SGPR32:#[0-9]+]]
+// CHECK: define amdgpu_kernel void @test_num_vgpr64_sgpr32() [[ATTR_VGPR64_SGPR32:#[0-9]+]]
 
 }
 
 __attribute__((amdgpu_num_sgpr(20), amdgpu_num_vgpr(40))) // expected-no-diagnostics
 kernel void test_num_sgpr20_vgpr40() {
-// CHECK: define void @test_num_sgpr20_vgpr40() [[ATTR_SGPR20_VGPR40:#[0-9]+]]
+// CHECK: define amdgpu_kernel void @test_num_sgpr20_vgpr40() [[ATTR_SGPR20_VGPR40:#[0-9]+]]
 }
 
 __attribute__((amdgpu_num_vgpr(0))) // expected-no-diagnostics
 kernel void test_num_vgpr0() {
 }
 
 __attribute__((amdgpu_num_sgpr(0))) // expected-no-diagnostics
 kernel void test_num_sgpr0() {