yaxunl updated this revision to Diff 205490.
yaxunl retitled this revision from "[HIP] Support texture type" to "[HIP]
Support device_shadow variable".
yaxunl edited the summary of this revision.
yaxunl added a comment.
Revised by Artem's comments.
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D62738/new/
https://reviews.llvm.org/D62738
Files:
include/clang/Basic/Attr.td
include/clang/Basic/AttrDocs.td
lib/CodeGen/CodeGenModule.cpp
lib/CodeGen/TargetInfo.cpp
lib/Sema/SemaDeclAttr.cpp
test/AST/ast-dump-cuda-device-shadow.cu
test/CodeGenCUDA/device-shadow.cu
test/Misc/pragma-attribute-supported-attributes-list.test
Index: test/Misc/pragma-attribute-supported-attributes-list.test
===================================================================
--- test/Misc/pragma-attribute-supported-attributes-list.test
+++ test/Misc/pragma-attribute-supported-attributes-list.test
@@ -26,6 +26,7 @@
// CHECK-NEXT: CPUSpecific (SubjectMatchRule_function)
// CHECK-NEXT: CUDAConstant (SubjectMatchRule_variable)
// CHECK-NEXT: CUDADevice (SubjectMatchRule_function, SubjectMatchRule_variable)
+// CHECK-NEXT: CUDADeviceShadow (SubjectMatchRule_variable)
// CHECK-NEXT: CUDAGlobal (SubjectMatchRule_function)
// CHECK-NEXT: CUDAHost (SubjectMatchRule_function)
// CHECK-NEXT: CUDALaunchBounds (SubjectMatchRule_objc_method, SubjectMatchRule_hasType_functionType)
Index: test/CodeGenCUDA/device-shadow.cu
===================================================================
--- /dev/null
+++ test/CodeGenCUDA/device-shadow.cu
@@ -0,0 +1,28 @@
+// REQUIRES: amdgpu-registered-target
+
+// RUN: %clang_cc1 -triple nvptx64-nvidia-cuda -std=c++11 -fcuda-is-device \
+// RUN: -emit-llvm -o - %s | FileCheck -check-prefixes=CUDADEV %s
+// RUN: %clang_cc1 -triple x86_64 -std=c++11 \
+// RUN: -emit-llvm -o - %s | FileCheck -check-prefixes=CUDAHOST %s
+
+// RUN: %clang_cc1 -triple amdgcn -fcuda-is-device -std=c++11 -fvisibility hidden -fapply-global-visibility-to-externs \
+// RUN: -emit-llvm -o - -x hip %s | FileCheck -check-prefixes=HIPDEV %s
+// RUN: %clang_cc1 -triple x86_64 -std=c++11 \
+// RUN: -emit-llvm -o - -x hip %s | FileCheck -check-prefixes=HIPHOST %s
+
+struct textureReference {
+ int a;
+};
+
+template <class T, int texType, int hipTextureReadMode>
+struct texture : public textureReference {
+texture() { a = 1; }
+};
+
+__attribute__((device_shadow)) texture<float, 2, 1> tex;
+// CUDADEV-NOT: @tex
+// CUDAHOST-NOT: call i32 @__hipRegisterVar{{.*}}@tex
+// HIPDEV: @tex = external protected{{.*}}global %struct.texture
+// HIPDEV-NOT: declare{{.*}}void @_ZN7textureIfLi2ELi1EEC1Ev
+// HIPHOST: define{{.*}}@_ZN7textureIfLi2ELi1EEC1Ev
+// HIPHOST: call i32 @__hipRegisterVar{{.*}}@tex{{.*}}i32 0, i32 4, i32 0, i32 0)
Index: test/AST/ast-dump-cuda-device-shadow.cu
===================================================================
--- /dev/null
+++ test/AST/ast-dump-cuda-device-shadow.cu
@@ -0,0 +1,13 @@
+// RUN: %clang_cc1 -fcuda-is-device -ast-dump -ast-dump-filter tex %s | FileCheck -strict-whitespace %s
+// RUN: %clang_cc1 -ast-dump -ast-dump-filter tex %s | FileCheck -strict-whitespace %s
+struct textureReference {
+ int a;
+};
+
+// CHECK: CUDADeviceShadowAttr
+template <class T, int texType, int hipTextureReadMode>
+struct texture : public textureReference {
+texture() { a = 1; }
+};
+
+__attribute__((device_shadow)) texture<float, 1, 1> tex;
Index: lib/Sema/SemaDeclAttr.cpp
===================================================================
--- lib/Sema/SemaDeclAttr.cpp
+++ lib/Sema/SemaDeclAttr.cpp
@@ -6786,6 +6786,10 @@
case ParsedAttr::AT_CUDAHost:
handleSimpleAttributeWithExclusions<CUDAHostAttr, CUDAGlobalAttr>(S, D, AL);
break;
+ case ParsedAttr::AT_CUDADeviceShadow:
+ handleSimpleAttributeWithExclusions<CUDADeviceShadowAttr, CUDADeviceAttr>(
+ S, D, AL);
+ break;
case ParsedAttr::AT_GNUInline:
handleGNUInlineAttr(S, D, AL);
break;
Index: lib/CodeGen/TargetInfo.cpp
===================================================================
--- lib/CodeGen/TargetInfo.cpp
+++ lib/CodeGen/TargetInfo.cpp
@@ -7848,7 +7848,8 @@
return D->hasAttr<OpenCLKernelAttr>() ||
(isa<FunctionDecl>(D) && D->hasAttr<CUDAGlobalAttr>()) ||
(isa<VarDecl>(D) &&
- (D->hasAttr<CUDADeviceAttr>() || D->hasAttr<CUDAConstantAttr>()));
+ (D->hasAttr<CUDADeviceAttr>() || D->hasAttr<CUDAConstantAttr>() ||
+ D->hasAttr<CUDADeviceShadowAttr>()));
}
void AMDGPUTargetCodeGenInfo::setTargetAttributes(
Index: lib/CodeGen/CodeGenModule.cpp
===================================================================
--- lib/CodeGen/CodeGenModule.cpp
+++ lib/CodeGen/CodeGenModule.cpp
@@ -2414,7 +2414,8 @@
if (!Global->hasAttr<CUDADeviceAttr>() &&
!Global->hasAttr<CUDAGlobalAttr>() &&
!Global->hasAttr<CUDAConstantAttr>() &&
- !Global->hasAttr<CUDASharedAttr>())
+ !Global->hasAttr<CUDASharedAttr>() &&
+ !(LangOpts.HIP && Global->hasAttr<CUDADeviceShadowAttr>()))
return;
} else {
// We need to emit host-side 'shadows' for all global
@@ -3769,7 +3770,12 @@
!getLangOpts().CUDAIsDevice &&
(D->hasAttr<CUDAConstantAttr>() || D->hasAttr<CUDADeviceAttr>() ||
D->hasAttr<CUDASharedAttr>());
- if (getLangOpts().CUDA && (IsCUDASharedVar || IsCUDAShadowVar))
+ // Device side shadow of initialized host-side global variables are also
+ // left undefined.
+ bool IsHIPDeviceShadowVar = getLangOpts().HIP && getLangOpts().CUDAIsDevice &&
+ D->hasAttr<CUDADeviceShadowAttr>();
+ if (getLangOpts().CUDA &&
+ (IsCUDASharedVar || IsCUDAShadowVar || IsHIPDeviceShadowVar))
Init = llvm::UndefValue::get(getTypes().ConvertType(ASTTy));
else if (!InitExpr) {
// This is a tentative definition; tentative definitions are
@@ -3880,7 +3886,8 @@
// global variables become internal definitions. These have to
// be internal in order to prevent name conflicts with global
// host variables with the same name in a different TUs.
- if (D->hasAttr<CUDADeviceAttr>() || D->hasAttr<CUDAConstantAttr>()) {
+ if (D->hasAttr<CUDADeviceAttr>() || D->hasAttr<CUDAConstantAttr>() ||
+ (D->hasAttr<CUDADeviceShadowAttr>() && LangOpts.HIP)) {
Linkage = llvm::GlobalValue::InternalLinkage;
// Shadow variables and their properties must be registered
@@ -3904,7 +3911,8 @@
}
}
- GV->setInitializer(Init);
+ if (!IsHIPDeviceShadowVar)
+ GV->setInitializer(Init);
if (emitter) emitter->finalize(GV);
// If it is safe to mark the global 'constant', do so now.
Index: include/clang/Basic/AttrDocs.td
===================================================================
--- include/clang/Basic/AttrDocs.td
+++ include/clang/Basic/AttrDocs.td
@@ -4157,3 +4157,17 @@
``__attribute__((malloc))``.
}];
}
+
+def DeviceShadowDocs : Documentation {
+ let Category = DocCatType;
+ let Content = [{
+The GNU style attribute __attribute__((device_shadow)) or MSVC style attribute
+__declspec(device_shadow) can be added to the definition of a global variable
+to indicate it is a HIP device shadow variable. A device shadow variable can
+be accessed on both device side and host side. It has external linkage and is
+not initialized on device side. It has internal linkage and is initialized by
+the initializer on host side.
+
+It is ignored for CUDA and other languages.
+ }];
+}
\ No newline at end of file
Index: include/clang/Basic/Attr.td
===================================================================
--- include/clang/Basic/Attr.td
+++ include/clang/Basic/Attr.td
@@ -951,6 +951,13 @@
let Documentation = [Undocumented];
}
+def CUDADeviceShadow : InheritableAttr {
+ let Spellings = [GNU<"device_shadow">, Declspec<"__device_shadow__">];
+ let Subjects = SubjectList<[Var]>;
+ let LangOpts = [CUDA];
+ let Documentation = [DeviceShadowDocs];
+}
+
def CUDADeviceBuiltin : IgnoredAttr {
let Spellings = [GNU<"device_builtin">, Declspec<"__device_builtin__">];
let LangOpts = [CUDA];
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
