Skip to content

[flang][cuda] Fix linkage for dynamic shared memory #161940

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.

Sign up for GitHub

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 1 commit into from
Oct 6, 2025
Merged

[flang][cuda] Fix linkage for dynamic shared memory #161940

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions flang/include/flang/Optimizer/Builder/FIRBuilder.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -372,6 +372,8 @@ class FirOpBuilder : public mlir::OpBuilder, public mlir::OpBuilder::Listener {
return createCommonLinkage(getContext());
}

mlir::StringAttr createExternalLinkage() { return getStringAttr("external"); }

mlir::StringAttr createInternalLinkage() { return getStringAttr("internal"); }

mlir::StringAttr createLinkOnceLinkage() { return getStringAttr("linkonce"); }
Expand Down
6 changes: 5 additions & 1 deletion flang/lib/Optimizer/Transforms/CUFComputeSharedMemoryOffsetsAndSize.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -143,7 +143,11 @@ struct CUFComputeSharedMemoryOffsetsAndSize
auto sharedMemType = fir::SequenceType::get(sharedMemSize, i8Ty);
std::string sharedMemGlobalName =
(funcOp.getName() + llvm::Twine(cudaSharedMemSuffix)).str();
mlir::StringAttr linkage = builder.createInternalLinkage();
// Dynamic shared memory needs an external linkage while static shared
// memory needs an internal linkage.
mlir::StringAttr linkage = nbDynamicSharedVariables > 0
? builder.createExternalLinkage()
: builder.createInternalLinkage();
builder.setInsertionPointToEnd(gpuMod.getBody());
llvm::SmallVector<mlir::NamedAttribute> attrs;
auto globalOpName = mlir::OperationName(fir::GlobalOp::getOperationName(),
Expand Down
4 changes: 3 additions & 1 deletion flang/test/Fir/CUDA/cuda-shared-offset.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,7 @@ module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<!llvm.ptr, dense<
// CHECK: %{{.*}} = cuf.shared_memory[%c0{{.*}} : i32] !fir.array<?xf32>, %c-1 : index {bindc_name = "r", uniq_name = "_QFdynsharedEr"} -> !fir.ref<!fir.array<?xf32>>
// CHECK: gpu.return
// CHECK: }
// CHECK: fir.global internal @_QPdynshared__shared_mem {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<0xi8>
// CHECK: fir.global external @_QPdynshared__shared_mem {alignment = 4 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<0xi8>

// -----

Expand Down Expand Up @@ -158,3 +158,5 @@ module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<!llvm.ptr, dense<
// CHECK-LABEL: gpu.func @_QMmtestsPtestany
// CHECK: %{{.*}} = cuf.shared_memory[%c0{{.*}} : i32] !fir.array<?xf64>, %c-1{{.*}} : index {bindc_name = "dmasks", uniq_name = "_QMmtestsFtestanyEdmasks"} -> !fir.ref<!fir.array<?xf64>>
// CHECK: %{{.*}} = cuf.shared_memory[%c0{{.*}} : i32] !fir.array<?xf32>, %c-1{{.*}} : index {bindc_name = "smasks", uniq_name = "_QMmtestsFtestanyEsmasks"} -> !fir.ref<!fir.array<?xf32>>

// CHECK: fir.global external @_QMmtestsPtestany__shared_mem {alignment = 8 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<0xi8>