Removed always-inline attribute from stubs and made required littest changes

Author: anikelal

clang/lib/AST/MicrosoftMangle.cpp

@@ -1163,7 +1163,7 @@ void MicrosoftCXXNameMangler::mangleUnqualifiedName(GlobalDecl GD,
                   ->hasAttr<CUDAGlobalAttr>())) &&
             GD.getKernelReferenceKind() == KernelReferenceKind::Stub;
         bool IsOCLDeviceStub =
-            ND && (isa<FunctionDecl>(ND) && ND->hasAttr<OpenCLKernelAttr>()) &&
+            ND && isa<FunctionDecl>(ND) && ND->hasAttr<OpenCLKernelAttr>() &&
             GD.getKernelReferenceKind() == KernelReferenceKind::Stub;
         if (IsDeviceStub)
           mangleSourceName(

clang/lib/CodeGen/CGCall.cpp

@@ -2504,7 +2504,12 @@ void CodeGenModule::ConstructAttributeList(StringRef Name,
                                  NumElemsParam);
     }
 
-    if (TargetDecl->hasAttr<OpenCLKernelAttr>()) {
+    if (TargetDecl->hasAttr<OpenCLKernelAttr>() &&
+        CallingConv != CallingConv::CC_C &&
+        CallingConv !=
+            CallingConv::CC_SpirFunction) { // Check CallingConv to avoid adding
+                                            // uniform-work-group-size attribute
+                                            // to OpenCL Kernel Stub
       if (getLangOpts().OpenCLVersion <= 120) {
         // OpenCL v1.2 Work groups are always uniform
         FuncAttrs.addAttribute("uniform-work-group-size", "true");

clang/lib/CodeGen/CGExpr.cpp

@@ -6144,11 +6144,9 @@ RValue CodeGenFunction::EmitCall(QualType CalleeType,
 
   const auto *FnType = cast<FunctionType>(PointeeType);
 
-  if (auto FD = dyn_cast_or_null<FunctionDecl>(TargetDecl)) {
-    if (FD->hasAttr<OpenCLKernelAttr>()) {
-      CGM.getTargetCodeGenInfo().setOCLKernelStubCallingConvention(FnType);
-    }
-  }
+  if (const auto *FD = dyn_cast_or_null<FunctionDecl>(TargetDecl);
+      FD && FD->hasAttr<OpenCLKernelAttr>())
+    CGM.getTargetCodeGenInfo().setOCLKernelStubCallingConvention(FnType);
 
   // If we are checking indirect calls and this call is indirect, check that the
   // function pointer is a member of the bit set for the function type.

clang/lib/CodeGen/CodeGenModule.cpp

@@ -6162,16 +6162,6 @@ void CodeGenModule::EmitGlobalFunctionDefinition(GlobalDecl GD,
 
   setNonAliasAttributes(GD, Fn);
 
-  if (D->hasAttr<OpenCLKernelAttr>()) {
-    if (GD.getKernelReferenceKind() == KernelReferenceKind::Stub) {
-      if (Fn->hasFnAttribute(llvm::Attribute::NoInline))
-        Fn->removeFnAttr(llvm::Attribute::NoInline);
-      if (Fn->hasFnAttribute(llvm::Attribute::InlineHint))
-        Fn->removeFnAttr(llvm::Attribute::InlineHint);
-      Fn->addFnAttr(llvm::Attribute::AlwaysInline);
-    }
-  }
-
   SetLLVMFunctionAttributesForDefinition(D, Fn);
 
   if (const ConstructorAttr *CA = D->getAttr<ConstructorAttr>())

clang/test/CodeGenOpenCL/addr-space-struct-arg.cl

@@ -106,7 +106,7 @@ kernel void test_target_features_kernel(global int *i) {
 // NOCPU-NEXT:    ret void
 //
 //
-// NOCPU: Function Attrs: alwaysinline convergent norecurse nounwind
+// NOCPU: Function Attrs: convergent noinline norecurse nounwind optnone
 // NOCPU-LABEL: define dso_local void @__clang_ocl_kern_imp_test(
 // NOCPU-SAME: ptr addrspace(1) noundef align 1 [[A:%.*]], i8 noundef signext [[B:%.*]], ptr addrspace(1) noundef align 8 [[C:%.*]], i64 noundef [[D:%.*]]) #[[ATTR3:[0-9]+]] !kernel_arg_addr_space [[META3]] !kernel_arg_access_qual [[META4]] !kernel_arg_type [[META5]] !kernel_arg_base_type [[META5]] !kernel_arg_type_qual [[META6]] {
 // NOCPU-NEXT:  [[ENTRY:.*:]]
@@ -230,7 +230,7 @@ kernel void test_target_features_kernel(global int *i) {
 // NOCPU-NEXT:    ret void
 //
 //
-// NOCPU: Function Attrs: alwaysinline convergent norecurse nounwind
+// NOCPU: Function Attrs: convergent noinline norecurse nounwind optnone
 // NOCPU-LABEL: define dso_local void @__clang_ocl_kern_imp_test_target_features_kernel(
 // NOCPU-SAME: ptr addrspace(1) noundef align 4 [[I:%.*]]) #[[ATTR5:[0-9]+]] !kernel_arg_addr_space [[META7]] !kernel_arg_access_qual [[META8]] !kernel_arg_type [[META9]] !kernel_arg_base_type [[META9]] !kernel_arg_type_qual [[META10]] {
 // NOCPU-NEXT:  [[ENTRY:.*:]]
@@ -362,7 +362,7 @@ kernel void test_target_features_kernel(global int *i) {
 // NOCPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[BLOCK_CAPTURE_ADDR]], align 8
 // NOCPU-NEXT:    [[BLOCK_CAPTURE_ADDR1:%.*]] = getelementptr inbounds nuw <{ i32, i32, ptr, i64, ptr addrspace(1) }>, ptr [[DOTBLOCK_DESCRIPTOR]], i32 0, i32 4
 // NOCPU-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr [[BLOCK_CAPTURE_ADDR1]], align 8
-// NOCPU-NEXT:    call void @callee(i64 noundef [[TMP0]], ptr addrspace(1) noundef [[TMP1]]) #[[ATTR11:[0-9]+]]
+// NOCPU-NEXT:    call void @callee(i64 noundef [[TMP0]], ptr addrspace(1) noundef [[TMP1]]) #[[ATTR10]]
 // NOCPU-NEXT:    ret void
 //
 //
@@ -451,7 +451,7 @@ kernel void test_target_features_kernel(global int *i) {
 // GFX900-NEXT:    ret void
 //
 //
-// GFX900: Function Attrs: alwaysinline convergent norecurse nounwind
+// GFX900: Function Attrs: convergent norecurse nounwind
 // GFX900-LABEL: define dso_local void @__clang_ocl_kern_imp_test(
 // GFX900-SAME: ptr addrspace(1) noundef align 1 [[A:%.*]], i8 noundef signext [[B:%.*]], ptr addrspace(1) noundef align 8 [[C:%.*]], i64 noundef [[D:%.*]]) #[[ATTR3:[0-9]+]] !kernel_arg_addr_space [[META10]] !kernel_arg_access_qual [[META11]] !kernel_arg_type [[META12]] !kernel_arg_base_type [[META12]] !kernel_arg_type_qual [[META13]] {
 // GFX900-NEXT:  [[ENTRY:.*:]]
@@ -585,7 +585,7 @@ kernel void test_target_features_kernel(global int *i) {
 // GFX900-NEXT:    ret void
 //
 //
-// GFX900: Function Attrs: alwaysinline convergent norecurse nounwind
+// GFX900: Function Attrs: convergent norecurse nounwind
 // GFX900-LABEL: define dso_local void @__clang_ocl_kern_imp_test_target_features_kernel(
 // GFX900-SAME: ptr addrspace(1) noundef align 4 [[I:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META22]] !kernel_arg_access_qual [[META23]] !kernel_arg_type [[META24]] !kernel_arg_base_type [[META24]] !kernel_arg_type_qual [[META25]] {
 // GFX900-NEXT:  [[ENTRY:.*:]]
@@ -715,7 +715,7 @@ kernel void test_target_features_kernel(global int *i) {
 // GFX900-NEXT:    [[TMP0:%.*]] = load i64, ptr [[BLOCK_CAPTURE_ADDR]], align 8, !tbaa [[TBAA3]]
 // GFX900-NEXT:    [[BLOCK_CAPTURE_ADDR1:%.*]] = getelementptr inbounds nuw <{ i32, i32, ptr, i64, ptr addrspace(1) }>, ptr [[DOTBLOCK_DESCRIPTOR]], i32 0, i32 4
 // GFX900-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr [[BLOCK_CAPTURE_ADDR1]], align 8, !tbaa [[TBAA7]]
-// GFX900-NEXT:    call void @callee(i64 noundef [[TMP0]], ptr addrspace(1) noundef [[TMP1]]) #[[ATTR10:[0-9]+]]
+// GFX900-NEXT:    call void @callee(i64 noundef [[TMP0]], ptr addrspace(1) noundef [[TMP1]]) #[[ATTR8]]
 // GFX900-NEXT:    ret void
 //
 //
@@ -754,27 +754,25 @@ kernel void test_target_features_kernel(global int *i) {
 // NOCPU: attributes #[[ATTR0:[0-9]+]] = { "objc_arc_inert" }
 // NOCPU: attributes #[[ATTR1]] = { convergent noinline norecurse nounwind optnone "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
 // NOCPU: attributes #[[ATTR2]] = { convergent noinline norecurse nounwind optnone "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
-// NOCPU: attributes #[[ATTR3]] = { alwaysinline convergent norecurse nounwind "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "uniform-work-group-size"="false" }
+// NOCPU: attributes #[[ATTR3]] = { convergent noinline norecurse nounwind optnone "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
 // NOCPU: attributes #[[ATTR4]] = { convergent noinline norecurse nounwind optnone "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+s-memtime-inst" "uniform-work-group-size"="false" }
-// NOCPU: attributes #[[ATTR5]] = { alwaysinline convergent norecurse nounwind "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+s-memtime-inst" "uniform-work-group-size"="false" }
+// NOCPU: attributes #[[ATTR5]] = { convergent noinline norecurse nounwind optnone "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+s-memtime-inst" }
 // NOCPU: attributes #[[ATTR6:[0-9]+]] = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }
 // NOCPU: attributes #[[ATTR7]] = { convergent noinline nounwind optnone "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
 // NOCPU: attributes #[[ATTR8]] = { convergent nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
 // NOCPU: attributes #[[ATTR9:[0-9]+]] = { nocallback nofree nosync nounwind willreturn }
-// NOCPU: attributes #[[ATTR10]] = { convergent nounwind "uniform-work-group-size"="false" }
-// NOCPU: attributes #[[ATTR11]] = { convergent nounwind }
+// NOCPU: attributes #[[ATTR10]] = { convergent nounwind }
 //.
 // GFX900: attributes #[[ATTR0:[0-9]+]] = { "objc_arc_inert" }
 // GFX900: attributes #[[ATTR1]] = { convergent norecurse nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64,-sram-ecc" }
 // GFX900: attributes #[[ATTR2]] = { convergent norecurse nounwind "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64,-sram-ecc" "uniform-work-group-size"="false" }
-// GFX900: attributes #[[ATTR3]] = { alwaysinline convergent norecurse nounwind "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64,-sram-ecc" "uniform-work-group-size"="false" }
+// GFX900: attributes #[[ATTR3]] = { convergent norecurse nounwind "amdgpu-flat-work-group-size"="1,256" "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64,-sram-ecc" }
 // GFX900: attributes #[[ATTR4:[0-9]+]] = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
 // GFX900: attributes #[[ATTR5:[0-9]+]] = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }
 // GFX900: attributes #[[ATTR6]] = { convergent nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx900" "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64,-sram-ecc" }
 // GFX900: attributes #[[ATTR7:[0-9]+]] = { nocallback nofree nosync nounwind willreturn }
-// GFX900: attributes #[[ATTR8]] = { convergent nounwind "uniform-work-group-size"="false" }
+// GFX900: attributes #[[ATTR8]] = { convergent nounwind }
 // GFX900: attributes #[[ATTR9]] = { nounwind }
-// GFX900: attributes #[[ATTR10]] = { convergent nounwind }
 //.
 // NOCPU: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 600}
 // NOCPU: [[META1:![0-9]+]] = !{i32 1, !"wchar_size", i32 4}

clang/test/CodeGenOpenCL/amdgpu-abi-struct-arg-byref.cl

@@ -5,6 +5,9 @@
 
 kernel void ker() {};
 // CHECK: define{{.*}}@ker() #[[ATTR0:[0-9]+]]
+// CHECK: call void @__clang_ocl_kern_imp_ker() #[[ATTR2:[0-9]+]]
+
+// CHECK: define{{.*}}@__clang_ocl_kern_imp_ker() #[[ATTR1:[0-9]+]]
 
 void foo() {};
 // CHECK: define{{.*}}@foo() #[[ATTR1:[0-9]+]]
@@ -15,3 +18,6 @@ void foo() {};
 
 // CHECK: attributes #[[ATTR1]]
 // CHECK-NOT: uniform-work-group-size
+
+// CHECK: attributes #[[ATTR2]]
+// CHECK-NOT: uniform-work-group-size