Skip to content

[flang][cuda] Support device component in a pointer or allocatable derived-type #149418

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
Jul 18, 2025
Merged

[flang][cuda] Support device component in a pointer or allocatable derived-type #149418

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
26 changes: 23 additions & 3 deletions flang/lib/Lower/ConvertVariable.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -814,6 +814,10 @@ initializeDeviceComponentAllocator(Fortran::lower::AbstractConverter &converter,
if (auto boxTy = mlir::dyn_cast<fir::BaseBoxType>(baseTy))
baseTy = boxTy.getEleTy();
baseTy = fir::unwrapRefType(baseTy);

if (mlir::isa<fir::SequenceType>(baseTy))
TODO(loc, "array of derived-type with device component");

auto recTy =
mlir::dyn_cast<fir::RecordType>(fir::unwrapSequenceType(baseTy));
assert(recTy && "expected fir::RecordType");
Expand All @@ -824,7 +828,7 @@ initializeDeviceComponentAllocator(Fortran::lower::AbstractConverter &converter,
if (Fortran::semantics::IsDeviceAllocatable(sym)) {
unsigned fieldIdx = recTy.getFieldIndex(sym.name().ToString());
mlir::Type fieldTy;
std::vector<mlir::Value> coordinates;
llvm::SmallVector<mlir::Value> coordinates;

if (fieldIdx != std::numeric_limits<unsigned>::max()) {
// Field found in the base record type.
Expand Down Expand Up @@ -867,8 +871,24 @@ initializeDeviceComponentAllocator(Fortran::lower::AbstractConverter &converter,
TODO(loc, "device resident component in complex derived-type "
"hierarchy");

mlir::Value comp = builder.create<fir::CoordinateOp>(
loc, builder.getRefType(fieldTy), fir::getBase(exv), coordinates);
mlir::Value base = fir::getBase(exv);
mlir::Value comp;
if (mlir::isa<fir::BaseBoxType>(fir::unwrapRefType(base.getType()))) {
mlir::Value box = builder.create<fir::LoadOp>(loc, base);
mlir::Value addr = builder.create<fir::BoxAddrOp>(loc, box);
llvm::SmallVector<mlir::Value> lenParams;
assert(coordinates.size() == 1 && "expect one coordinate");
auto field = mlir::dyn_cast<fir::FieldIndexOp>(
coordinates[0].getDefiningOp());
comp = builder.create<hlfir::DesignateOp>(
loc, builder.getRefType(fieldTy), addr,
/*component=*/field.getFieldName(),
/*componentShape=*/mlir::Value{},
hlfir::DesignateOp::Subscripts{});
} else {
comp = builder.create<fir::CoordinateOp>(
loc, builder.getRefType(fieldTy), base, coordinates);
}
cuf::DataAttributeAttr dataAttr =
Fortran::lower::translateSymbolCUFDataAttribute(
builder.getContext(), sym);
Expand Down
32 changes: 32 additions & 0 deletions flang/test/Lower/CUDA/cuda-set-allocator.cuf
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,4 +21,36 @@ contains
! CHECK: %[[Z:.*]] = fir.coordinate_of %[[DT]]#0, 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>}

subroutine sub2()
type(ty_device), pointer :: d1
end subroutine

! CHECK-LABEL: func.func @_QMm1Psub2()
! CHECK: %[[ALLOC:.*]] = cuf.alloc !fir.box<!fir.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>> {bindc_name = "d1", data_attr = #cuf.cuda<managed>, uniq_name = "_QMm1Fsub2Ed1"} -> !fir.ref<!fir.box<!fir.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>>
! CHECK: %[[DECL:.*]]:2 = hlfir.declare %[[ALLOC]] {data_attr = #cuf.cuda<managed>, fortran_attrs = #fir.var_attrs<pointer>, uniq_name = "_QMm1Fsub2Ed1"} : (!fir.ref<!fir.box<!fir.ptr<!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.ptr<!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.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>>)
! CHECK: %[[LOAD1:.*]] = fir.load %[[DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>>
! CHECK: %[[ADDR1:.*]] = fir.box_addr %[[LOAD1]] : (!fir.box<!fir.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>) -> !fir.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>
! CHECK: %[[DESIGNATE1:.*]] = hlfir.designate %[[ADDR1]]{"x"} : (!fir.ptr<!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 %[[DESIGNATE1]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>> {data_attr = #cuf.cuda<device>}
! CHECK: %[[LOAD2:.*]] = fir.load %[[DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>>
! CHECK: %[[ADDR2:.*]] = fir.box_addr %[[LOAD2]] : (!fir.box<!fir.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>) -> !fir.ptr<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>
! CHECK: %[[DESIGNATE2:.*]] = hlfir.designate %[[ADDR2]]{"z"} : (!fir.ptr<!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 %[[DESIGNATE2]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>> {data_attr = #cuf.cuda<device>}

subroutine sub3()
type(ty_device), allocatable :: d1
end subroutine

! CHECK-LABEL: func.func @_QMm1Psub3()
! CHECK: %[[ALLOC:.*]] = cuf.alloc !fir.box<!fir.heap<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>> {bindc_name = "d1", data_attr = #cuf.cuda<managed>, uniq_name = "_QMm1Fsub3Ed1"} -> !fir.ref<!fir.box<!fir.heap<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>>
! CHECK: %[[DECL:.*]]:2 = hlfir.declare %[[ALLOC]] {data_attr = #cuf.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QMm1Fsub3Ed1"} : (!fir.ref<!fir.box<!fir.heap<!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.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.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>>)
! CHECK: %[[LOAD1:.*]] = fir.load %[[DECL]]#0 : !fir.ref<!fir.box<!fir.heap<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>>
! CHECK: %[[ADDR1:.*]] = fir.box_addr %[[LOAD1]] : (!fir.box<!fir.heap<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>) -> !fir.heap<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>
! CHECK: %[[DESIGNATE1:.*]] = hlfir.designate %[[ADDR1]]{"x"} : (!fir.heap<!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 %[[DESIGNATE1]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>> {data_attr = #cuf.cuda<device>}
! CHECK: %[[LOAD2:.*]] = fir.load %[[DECL]]#0 : !fir.ref<!fir.box<!fir.heap<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>>
! CHECK: %[[ADDR2:.*]] = fir.box_addr %[[LOAD2]] : (!fir.box<!fir.heap<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>>) -> !fir.heap<!fir.type<_QMm1Tty_device{x:!fir.box<!fir.heap<!fir.array<?xi32>>>,y:i32,z:!fir.box<!fir.heap<!fir.array<?xi32>>>}>>
! CHECK: %[[DESIGNATE2:.*]] = hlfir.designate %[[ADDR2]]{"z"} : (!fir.heap<!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 %[[DESIGNATE2]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>> {data_attr = #cuf.cuda<device>}

end module
Loading