[CUDA][HIP][NFC] add CodeGenModule::shouldEmitCUDAGlobalVar #98543
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Extract the logic whether to emit a global var based on CUDA/HIP host/device related attributes to CodeGenModule::shouldEmitCUDAGlobalVar to be used by other places.
llvmbot
added
clang
|
@llvm/pr-subscribers-clang-codegen Author: Yaxun (Sam) Liu (yxsamliu) ChangesExtract the logic whether to emit a global var based on CUDA/HIP host/device related attributes to CodeGenModule::shouldEmitCUDAGlobalVar to be used by other places. Full diff: https://github.com/llvm/llvm-project/pull/98543.diff 2 Files Affected:
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 5c810cd332185..6c10b4a2edef8 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -3702,6 +3702,19 @@ template <typename AttrT> static bool hasImplicitAttr(const ValueDecl *D) {
return D->isImplicit();
}
+bool CodeGenModule::shouldEmitCUDAGlobalVar(const VarDecl *Global) const {
+ assert(LangOpts.CUDA && "Should not be called by non-CUDA languages");
+ // We need to emit host-side 'shadows' for all global
+ // device-side variables because the CUDA runtime needs their
+ // size and host-side address in order to provide access to
+ // their device-side incarnations.
+ return !LangOpts.CUDAIsDevice || Global->hasAttr<CUDADeviceAttr>() ||
+ Global->hasAttr<CUDAConstantAttr>() ||
+ Global->hasAttr<CUDASharedAttr>() ||
+ Global->getType()->isCUDADeviceBuiltinSurfaceType() ||
+ Global->getType()->isCUDADeviceBuiltinTextureType();
+}
+
void CodeGenModule::EmitGlobal(GlobalDecl GD) {
const auto *Global = cast<ValueDecl>(GD.getDecl());
@@ -3726,36 +3739,27 @@ void CodeGenModule::EmitGlobal(GlobalDecl GD) {
// Non-constexpr non-lambda implicit host device functions are not emitted
// unless they are used on device side.
if (LangOpts.CUDA) {
- if (LangOpts.CUDAIsDevice) {
+ assert((isa<FunctionDecl>(Global) || isa<VarDecl>(Global)) &&
+ "Expected Variable or Function");
+ if (const auto *VD = dyn_cast<VarDecl>(Global)) {
+ if (!shouldEmitCUDAGlobalVar(VD))
+ return;
+ } else if (LangOpts.CUDAIsDevice) {
const auto *FD = dyn_cast<FunctionDecl>(Global);
if ((!Global->hasAttr<CUDADeviceAttr>() ||
- (LangOpts.OffloadImplicitHostDeviceTemplates && FD &&
+ (LangOpts.OffloadImplicitHostDeviceTemplates &&
hasImplicitAttr<CUDAHostAttr>(FD) &&
hasImplicitAttr<CUDADeviceAttr>(FD) && !FD->isConstexpr() &&
!isLambdaCallOperator(FD) &&
!getContext().CUDAImplicitHostDeviceFunUsedByDevice.count(FD))) &&
!Global->hasAttr<CUDAGlobalAttr>() &&
- !Global->hasAttr<CUDAConstantAttr>() &&
- !Global->hasAttr<CUDASharedAttr>() &&
- !Global->getType()->isCUDADeviceBuiltinSurfaceType() &&
- !Global->getType()->isCUDADeviceBuiltinTextureType() &&
!(LangOpts.HIPStdPar && isa<FunctionDecl>(Global) &&
!Global->hasAttr<CUDAHostAttr>()))
return;
- } else {
- // We need to emit host-side 'shadows' for all global
- // device-side variables because the CUDA runtime needs their
- // size and host-side address in order to provide access to
- // their device-side incarnations.
-
- // So device-only functions are the only things we skip.
- if (isa<FunctionDecl>(Global) && !Global->hasAttr<CUDAHostAttr>() &&
- Global->hasAttr<CUDADeviceAttr>())
- return;
-
- assert((isa<FunctionDecl>(Global) || isa<VarDecl>(Global)) &&
- "Expected Variable or Function");
- }
+ // Device-only functions are the only things we skip.
+ } else if (!Global->hasAttr<CUDAHostAttr>() &&
+ Global->hasAttr<CUDADeviceAttr>())
+ return;
}
if (LangOpts.OpenMP) {
diff --git a/clang/lib/CodeGen/CodeGenModule.h b/clang/lib/CodeGen/CodeGenModule.h
index 8b65348b879b6..392e2bf0cdfea 100644
--- a/clang/lib/CodeGen/CodeGenModule.h
+++ b/clang/lib/CodeGen/CodeGenModule.h
@@ -563,6 +563,9 @@ class CodeGenModule : public CodeGenTypeCache {
bool isTriviallyRecursive(const FunctionDecl *F);
bool shouldEmitFunction(GlobalDecl GD);
+ // Whether a global variable should be emitted by CUDA/HIP host/device
+ // related attributes.
+ bool shouldEmitCUDAGlobalVar(const VarDecl *VD) const;
bool shouldOpportunisticallyEmitVTables();
/// Map used to be sure we don't emit the same CompoundLiteral twice.
llvm::DenseMap<const CompoundLiteralExpr *, llvm::GlobalVariable *>
|
|
@llvm/pr-subscribers-clang Author: Yaxun (Sam) Liu (yxsamliu) ChangesExtract the logic whether to emit a global var based on CUDA/HIP host/device related attributes to CodeGenModule::shouldEmitCUDAGlobalVar to be used by other places. Full diff: https://github.com/llvm/llvm-project/pull/98543.diff 2 Files Affected:
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 5c810cd332185..6c10b4a2edef8 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -3702,6 +3702,19 @@ template <typename AttrT> static bool hasImplicitAttr(const ValueDecl *D) {
return D->isImplicit();
}
+bool CodeGenModule::shouldEmitCUDAGlobalVar(const VarDecl *Global) const {
+ assert(LangOpts.CUDA && "Should not be called by non-CUDA languages");
+ // We need to emit host-side 'shadows' for all global
+ // device-side variables because the CUDA runtime needs their
+ // size and host-side address in order to provide access to
+ // their device-side incarnations.
+ return !LangOpts.CUDAIsDevice || Global->hasAttr<CUDADeviceAttr>() ||
+ Global->hasAttr<CUDAConstantAttr>() ||
+ Global->hasAttr<CUDASharedAttr>() ||
+ Global->getType()->isCUDADeviceBuiltinSurfaceType() ||
+ Global->getType()->isCUDADeviceBuiltinTextureType();
+}
+
void CodeGenModule::EmitGlobal(GlobalDecl GD) {
const auto *Global = cast<ValueDecl>(GD.getDecl());
@@ -3726,36 +3739,27 @@ void CodeGenModule::EmitGlobal(GlobalDecl GD) {
// Non-constexpr non-lambda implicit host device functions are not emitted
// unless they are used on device side.
if (LangOpts.CUDA) {
- if (LangOpts.CUDAIsDevice) {
+ assert((isa<FunctionDecl>(Global) || isa<VarDecl>(Global)) &&
+ "Expected Variable or Function");
+ if (const auto *VD = dyn_cast<VarDecl>(Global)) {
+ if (!shouldEmitCUDAGlobalVar(VD))
+ return;
+ } else if (LangOpts.CUDAIsDevice) {
const auto *FD = dyn_cast<FunctionDecl>(Global);
if ((!Global->hasAttr<CUDADeviceAttr>() ||
- (LangOpts.OffloadImplicitHostDeviceTemplates && FD &&
+ (LangOpts.OffloadImplicitHostDeviceTemplates &&
hasImplicitAttr<CUDAHostAttr>(FD) &&
hasImplicitAttr<CUDADeviceAttr>(FD) && !FD->isConstexpr() &&
!isLambdaCallOperator(FD) &&
!getContext().CUDAImplicitHostDeviceFunUsedByDevice.count(FD))) &&
!Global->hasAttr<CUDAGlobalAttr>() &&
- !Global->hasAttr<CUDAConstantAttr>() &&
- !Global->hasAttr<CUDASharedAttr>() &&
- !Global->getType()->isCUDADeviceBuiltinSurfaceType() &&
- !Global->getType()->isCUDADeviceBuiltinTextureType() &&
!(LangOpts.HIPStdPar && isa<FunctionDecl>(Global) &&
!Global->hasAttr<CUDAHostAttr>()))
return;
- } else {
- // We need to emit host-side 'shadows' for all global
- // device-side variables because the CUDA runtime needs their
- // size and host-side address in order to provide access to
- // their device-side incarnations.
-
- // So device-only functions are the only things we skip.
- if (isa<FunctionDecl>(Global) && !Global->hasAttr<CUDAHostAttr>() &&
- Global->hasAttr<CUDADeviceAttr>())
- return;
-
- assert((isa<FunctionDecl>(Global) || isa<VarDecl>(Global)) &&
- "Expected Variable or Function");
- }
+ // Device-only functions are the only things we skip.
+ } else if (!Global->hasAttr<CUDAHostAttr>() &&
+ Global->hasAttr<CUDADeviceAttr>())
+ return;
}
if (LangOpts.OpenMP) {
diff --git a/clang/lib/CodeGen/CodeGenModule.h b/clang/lib/CodeGen/CodeGenModule.h
index 8b65348b879b6..392e2bf0cdfea 100644
--- a/clang/lib/CodeGen/CodeGenModule.h
+++ b/clang/lib/CodeGen/CodeGenModule.h
@@ -563,6 +563,9 @@ class CodeGenModule : public CodeGenTypeCache {
bool isTriviallyRecursive(const FunctionDecl *F);
bool shouldEmitFunction(GlobalDecl GD);
+ // Whether a global variable should be emitted by CUDA/HIP host/device
+ // related attributes.
+ bool shouldEmitCUDAGlobalVar(const VarDecl *VD) const;
bool shouldOpportunisticallyEmitVTables();
/// Map used to be sure we don't emit the same CompoundLiteral twice.
llvm::DenseMap<const CompoundLiteralExpr *, llvm::GlobalVariable *>
|
Artem-B
approved these changes
Jul 11, 2024
aaryanshukla
pushed a commit
to aaryanshukla/llvm-project
that referenced
this pull request
Jul 14, 2024
Extract the logic whether to emit a global var based on CUDA/HIP host/device related attributes to CodeGenModule::shouldEmitCUDAGlobalVar to be used by other places.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Extract the logic whether to emit a global var based on CUDA/HIP host/device related attributes to CodeGenModule::shouldEmitCUDAGlobalVar to be used by other places.