Skip to content

[flang][cuda] Lower CUDA shared variable with cuf.shared_memory op #131399

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
clementval merged 2 commits into from
Mar 17, 2025
Merged

[flang][cuda] Lower CUDA shared variable with cuf.shared_memory op #131399

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
74b42bb
[flang][cuda] Lower CUDA shared variable with cuf.shared_memory opera…
clementval Mar 14, 2025
2fa44f0
Add check-not for cuf.free
clementval Mar 16, 2025
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
14 changes: 10 additions & 4 deletions flang/lib/Lower/ConvertVariable.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -738,9 +738,11 @@ static mlir::Value createNewLocal(Fortran::lower::AbstractConverter &converter,
auto idxTy = builder.getIndexType();
for (mlir::Value sh : elidedShape)
indices.push_back(builder.createConvert(loc, idxTy, sh));
mlir::Value alloc = builder.create<cuf::AllocOp>(
loc, ty, nm, symNm, dataAttr, lenParams, indices);
return alloc;
if (dataAttr.getValue() == cuf::DataAttribute::Shared)
return builder.create<cuf::SharedMemoryOp>(loc, ty, nm, symNm, lenParams,
indices);
return builder.create<cuf::AllocOp>(loc, ty, nm, symNm, dataAttr, lenParams,
indices);
}

// Let the builder do all the heavy lifting.
Expand Down Expand Up @@ -1032,12 +1034,16 @@ static void instantiateLocal(Fortran::lower::AbstractConverter &converter,
symMap);
if (Fortran::semantics::NeedCUDAAlloc(var.getSymbol())) {
auto *builder = &converter.getFirOpBuilder();
cuf::DataAttributeAttr dataAttr =
Fortran::lower::translateSymbolCUFDataAttribute(builder->getContext(),
var.getSymbol());
mlir::Location loc = converter.getCurrentLocation();
fir::ExtendedValue exv =
converter.getSymbolExtendedValue(var.getSymbol(), &symMap);
auto *sym = &var.getSymbol();
const Fortran::semantics::Scope &owner = sym->owner();
if (owner.kind() != Fortran::semantics::Scope::Kind::MainProgram) {
if (owner.kind() != Fortran::semantics::Scope::Kind::MainProgram &&
dataAttr.getValue() != cuf::DataAttribute::Shared) {
converter.getFctCtx().attachCleanup([builder, loc, exv, sym]() {
cuf::DataAttributeAttr dataAttr =
Fortran::lower::translateSymbolCUFDataAttribute(
Expand Down
11 changes: 11 additions & 0 deletions flang/test/Lower/CUDA/cuda-shared.cuf
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
! RUN: bbc -emit-hlfir -fcuda %s -o - | FileCheck %s

attributes(global) subroutine sharedmem()
real, shared :: s(32)
integer :: t
t = threadIdx%x
s(t) = t
end subroutine

! CHECK-LABEL: func.func @_QPsharedmem() attributes {cuf.proc_attr = #cuf.cuda_proc<global>}
! CHECK: %{{.*}} = cuf.shared_memory !fir.array<32xf32> {bindc_name = "s", uniq_name = "_QFsharedmemEs"} -> !fir.ref<!fir.array<32xf32>>