Skip to content

[flang][cuda] Pass descriptor by reference for CUFMemsetDescriptor #114338

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 31, 2024
Merged

[flang][cuda] Pass descriptor by reference for CUFMemsetDescriptor #114338

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: 1 addition & 1 deletion flang/include/flang/Runtime/CUDA/memory.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,7 @@ void RTDECL(CUFMemFree)(void *devicePtr, unsigned type,
/// Set value to the data hold by a descriptor. The \p value pointer must be
/// addressable to the same amount of bytes specified by the element size of
/// the descriptor \p desc.
void RTDECL(CUFMemsetDescriptor)(const Descriptor &desc, void *value,
void RTDECL(CUFMemsetDescriptor)(Descriptor *desc, void *value,
const char *sourceFile = nullptr, int sourceLine = 0);

/// Data transfer from a pointer to a pointer.
Expand Down
3 changes: 1 addition & 2 deletions flang/lib/Optimizer/Transforms/CUFOpConversion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -552,9 +552,8 @@ struct CUFDataTransferOpConversion
mlir::Value sourceFile = fir::factory::locationToFilename(builder, loc);
mlir::Value sourceLine =
fir::factory::locationToLineNo(builder, loc, fTy.getInput(3));
mlir::Value dst = builder.loadIfRef(loc, op.getDst());
llvm::SmallVector<mlir::Value> args{fir::runtime::createArguments(
builder, loc, fTy, dst, val, sourceFile, sourceLine)};
builder, loc, fTy, op.getDst(), val, sourceFile, sourceLine)};
builder.create<fir::CallOp>(loc, func, args);
rewriter.eraseOp(op);
} else {
Expand Down
4 changes: 2 additions & 2 deletions flang/runtime/CUDA/memory.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,8 +49,8 @@ void RTDEF(CUFMemFree)(
}
}

void RTDEF(CUFMemsetDescriptor)(const Descriptor &desc, void *value,
const char *sourceFile, int sourceLine) {
void RTDEF(CUFMemsetDescriptor)(
Descriptor *desc, void *value, const char *sourceFile, int sourceLine) {
Terminator terminator{sourceFile, sourceLine};
terminator.Crash("not yet implemented: CUDA data transfer from a scalar "
"value to a descriptor");
Expand Down
10 changes: 4 additions & 6 deletions flang/test/Fir/CUDA/cuda-data-transfer.fir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,10 +33,9 @@ func.func @_QPsub2() {
// CHECK: %[[ADEV:.*]]:2 = hlfir.declare %{{.*}} {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub2Eadev"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>)
// CHECK: %[[C2:.*]] = arith.constant 2 : i32
// CHECK: fir.store %[[C2]] to %[[TEMP]] : !fir.ref<i32>
// CHECK: %[[ADEV_LOAD:.*]] = fir.load %[[ADEV]]#0 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
// CHECK: %[[ADEV_BOX:.*]] = fir.convert %[[ADEV_LOAD]] : (!fir.box<!fir.heap<!fir.array<?xi32>>>) -> !fir.box<none>
// CHECK: %[[ADEV_BOX:.*]] = fir.convert %[[ADEV]]#0 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>) -> !fir.ref<!fir.box<none>>
// CHECK: %[[TEMP_CONV:.*]] = fir.convert %[[TEMP]] : (!fir.ref<i32>) -> !fir.llvm_ptr<i8>
// CHECK: fir.call @_FortranACUFMemsetDescriptor(%[[ADEV_BOX]], %[[TEMP_CONV]], %{{.*}}, %{{.*}}) : (!fir.box<none>, !fir.llvm_ptr<i8>, !fir.ref<i8>, i32) -> none
// CHECK: fir.call @_FortranACUFMemsetDescriptor(%[[ADEV_BOX]], %[[TEMP_CONV]], %{{.*}}, %{{.*}}) : (!fir.ref<!fir.box<none>>, !fir.llvm_ptr<i8>, !fir.ref<i8>, i32) -> none

func.func @_QPsub3() {
%0 = cuf.alloc !fir.box<!fir.heap<!fir.array<?xi32>>> {bindc_name = "adev", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub3Eadev"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
Expand All @@ -51,10 +50,9 @@ func.func @_QPsub3() {
// CHECK-LABEL: func.func @_QPsub3()
// CHECK: %[[ADEV:.*]]:2 = hlfir.declare %{{.*}} {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub3Eadev"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>)
// CHECK: %[[V:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsub3Ev"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
// CHECK: %[[ADEV_LOAD:.*]] = fir.load %[[ADEV]]#0 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
// CHECK: %[[ADEV_BOX:.*]] = fir.convert %[[ADEV_LOAD]] : (!fir.box<!fir.heap<!fir.array<?xi32>>>) -> !fir.box<none>
// CHECK: %[[ADEV_BOX:.*]] = fir.convert %[[ADEV]]#0 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>) -> !fir.ref<!fir.box<none>>
// CHECK: %[[V_CONV:.*]] = fir.convert %[[V]]#0 : (!fir.ref<i32>) -> !fir.llvm_ptr<i8>
// CHECK: fir.call @_FortranACUFMemsetDescriptor(%[[ADEV_BOX]], %[[V_CONV]], %{{.*}}, %{{.*}}) : (!fir.box<none>, !fir.llvm_ptr<i8>, !fir.ref<i8>, i32) -> none
// CHECK: fir.call @_FortranACUFMemsetDescriptor(%[[ADEV_BOX]], %[[V_CONV]], %{{.*}}, %{{.*}}) : (!fir.ref<!fir.box<none>>, !fir.llvm_ptr<i8>, !fir.ref<i8>, i32) -> none

func.func @_QPsub4() {
%0 = cuf.alloc !fir.box<!fir.heap<!fir.array<?xi32>>> {bindc_name = "adev", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub4Eadev"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
Expand Down
Loading