[flang][cuda] Add cuf.set_allocator_idx for device component #148750
Merged
Conversation
This file contains hidden or 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
llvmbot
added
flang
|
@llvm/pr-subscribers-flang-semantics @llvm/pr-subscribers-flang-fir-hlfir Author: Valentin Clement (バレンタイン クレメン) (clementval) ChangesUse the Full diff: https://github.com/llvm/llvm-project/pull/148750.diff 3 Files Affected:
diff --git a/flang/include/flang/Semantics/tools.h b/flang/include/flang/Semantics/tools.h
index fb670528f3ce4..317b9357b4c1f 100644
--- a/flang/include/flang/Semantics/tools.h
+++ b/flang/include/flang/Semantics/tools.h
@@ -199,6 +199,8 @@ bool IsPolymorphic(const Symbol &);
bool IsUnlimitedPolymorphic(const Symbol &);
bool IsPolymorphicAllocatable(const Symbol &);
+bool IsDeviceAllocatable(const Symbol &symbol);
+
inline bool IsCUDADeviceContext(const Scope *scope) {
if (scope) {
if (const Symbol * symbol{scope->symbol()}) {
diff --git a/flang/lib/Lower/ConvertVariable.cpp b/flang/lib/Lower/ConvertVariable.cpp
index 44f534e7d569a..cb931895600df 100644
--- a/flang/lib/Lower/ConvertVariable.cpp
+++ b/flang/lib/Lower/ConvertVariable.cpp
@@ -771,9 +771,79 @@ static mlir::Value createNewLocal(Fortran::lower::AbstractConverter &converter,
return builder.create<cuf::SharedMemoryOp>(loc, ty, nm, symNm, lenParams,
indices);
- if (!cuf::isCUDADeviceContext(builder.getRegion()))
- return builder.create<cuf::AllocOp>(loc, ty, nm, symNm, dataAttr,
- lenParams, indices);
+ if (!cuf::isCUDADeviceContext(builder.getRegion())) {
+ mlir::Value alloc = builder.create<cuf::AllocOp>(
+ loc, ty, nm, symNm, dataAttr, lenParams, indices);
+ if (const auto *details{
+ ultimateSymbol
+ .detailsIf<Fortran::semantics::ObjectEntityDetails>()}) {
+ const Fortran::semantics::DeclTypeSpec *type{details->type()};
+ const Fortran::semantics::DerivedTypeSpec *derived{
+ type ? type->AsDerived() : nullptr};
+ if (derived) {
+ Fortran::semantics::UltimateComponentIterator components{*derived};
+ auto recTy = mlir::dyn_cast<fir::RecordType>(ty);
+
+ mlir::Type fieldTy;
+ llvm::SmallVector<mlir::Value> coordinates;
+ for (const auto &sym : components) {
+ if (Fortran::semantics::IsDeviceAllocatable(sym)) {
+ unsigned fieldIdx = recTy.getFieldIndex(sym.name().ToString());
+ mlir::Type fieldTy;
+ std::vector<mlir::Value> coordinates;
+
+ if (fieldIdx != std::numeric_limits<unsigned>::max()) {
+ // Field found in the base record type.
+ auto fieldName = recTy.getTypeList()[fieldIdx].first;
+ fieldTy = recTy.getTypeList()[fieldIdx].second;
+ mlir::Value fieldIndex = builder.create<fir::FieldIndexOp>(
+ loc, fir::FieldType::get(fieldTy.getContext()), fieldName,
+ recTy,
+ /*typeParams=*/mlir::ValueRange{});
+ coordinates.push_back(fieldIndex);
+ } else {
+ // Field not found in base record type, search in potential
+ // record type components.
+ for (auto component : recTy.getTypeList()) {
+ if (auto childRecTy =
+ mlir::dyn_cast<fir::RecordType>(component.second)) {
+ fieldIdx = childRecTy.getFieldIndex(sym.name().ToString());
+ if (fieldIdx != std::numeric_limits<unsigned>::max()) {
+ mlir::Value parentFieldIndex =
+ builder.create<fir::FieldIndexOp>(
+ loc, fir::FieldType::get(childRecTy.getContext()),
+ component.first, recTy,
+ /*typeParams=*/mlir::ValueRange{});
+ coordinates.push_back(parentFieldIndex);
+ auto fieldName = childRecTy.getTypeList()[fieldIdx].first;
+ fieldTy = childRecTy.getTypeList()[fieldIdx].second;
+ mlir::Value childFieldIndex =
+ builder.create<fir::FieldIndexOp>(
+ loc, fir::FieldType::get(fieldTy.getContext()),
+ fieldName, childRecTy,
+ /*typeParams=*/mlir::ValueRange{});
+ coordinates.push_back(childFieldIndex);
+ break;
+ }
+ }
+ }
+ }
+
+ if (coordinates.empty())
+ TODO(loc, "device resident component in complex derived-type hierarchy");
+
+ mlir::Value comp = builder.create<fir::CoordinateOp>(
+ loc, builder.getRefType(fieldTy), alloc, coordinates);
+ cuf::DataAttributeAttr dataAttr =
+ Fortran::lower::translateSymbolCUFDataAttribute(
+ builder.getContext(), sym);
+ builder.create<cuf::SetAllocatorIndexOp>(loc, comp, dataAttr);
+ }
+ }
+ }
+ }
+ return alloc;
+ }
}
// Let the builder do all the heavy lifting.
diff --git a/flang/test/Lower/CUDA/cuda-set-allocator.cuf b/flang/test/Lower/CUDA/cuda-set-allocator.cuf
new file mode 100644
index 0000000000000..bf74e012a639d
--- /dev/null
+++ b/flang/test/Lower/CUDA/cuda-set-allocator.cuf
@@ -0,0 +1,21 @@
+! RUN: bbc -emit-hlfir -fcuda %s -o - | FileCheck %s
+
+module m1
+ type ty_device
+ integer, device, allocatable, dimension(:) :: x
+ integer :: y
+ integer, device, allocatable, dimension(:) :: z
+ end type
+contains
+ subroutine sub1()
+ type(ty_device) :: a
+ end subroutine
+
+! CHECK-LABEL: func.func @_QMm1Psub1()
+! CHECK: %[[DT:.*]] = cuf.alloc !fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}> {bindc_name = "a", data_attr = #cuf.cuda<managed>, uniq_name = "_QMm1Fsub1Ea"} -> !fir.ref<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>
+! CHECK: %[[X:.*]] = fir.coordinate_of %[[DT]], x : (!fir.ref<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
+! CHECK: cuf.set_allocator_idx %[[X]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>> {data_attr = #cuf.cuda<device>}
+! CHECK: %[[Z:.*]] = fir.coordinate_of %[[DT]], z : (!fir.ref<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
+! CHECK: cuf.set_allocator_idx %[[Z]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>> {data_attr = #cuf.cuda<device>}
+
+end module
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
wangzpgi
approved these changes
Jul 15, 2025
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/157/builds/33785 Here is the relevant piece of the build log for the reference |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
Use the
cuf.set_allocator_idxoperation introduced in #148717 when allocating a derived-type with device resident components.