Skip to content
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

Jump to bottom

[flang][cuda] Use cuf.alloc/cuf.free for local descriptor #98518

Merged
merged 3 commits into from
Jul 17, 2024
Merged

[flang][cuda] Use cuf.alloc/cuf.free for local descriptor #98518

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
3053909
[flang][cuda] Use cuf.alloc/cuf.free for local descriptor holding cud…
clementval Jul 11, 2024
4b0b77f
Make sure cuf.free is placed after conditional deallocation of the data
clementval Jul 11, 2024
3fc607c
Remove commented code
clementval Jul 11, 2024
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
8 changes: 2 additions & 6 deletions flang/include/flang/Semantics/tools.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -222,19 +222,15 @@ inline bool HasCUDAAttr(const Symbol &sym) {
}

inline bool NeedCUDAAlloc(const Symbol &sym) {
bool inDeviceSubprogram{IsCUDADeviceContext(&sym.owner())};
if (IsDummy(sym)) {
return false;
}
if (const auto *details{sym.GetUltimate().detailsIf<ObjectEntityDetails>()}) {
if (details->cudaDataAttr() &&
(*details->cudaDataAttr() == common::CUDADataAttr::Device ||
*details->cudaDataAttr() == common::CUDADataAttr::Managed ||
*details->cudaDataAttr() == common::CUDADataAttr::Unified)) {
// Descriptor is allocated on host when in host context.
if (IsAllocatable(sym)) {
return inDeviceSubprogram;
}
*details->cudaDataAttr() == common::CUDADataAttr::Unified ||
*details->cudaDataAttr() == common::CUDADataAttr::Pinned)) {
return true;
}
}
Expand Down
32 changes: 17 additions & 15 deletions flang/lib/Lower/ConvertVariable.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -715,8 +715,9 @@ 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));
return builder.create<cuf::AllocOp>(loc, ty, nm, symNm, dataAttr, lenParams,
indices);
mlir::Value alloc = builder.create<cuf::AllocOp>(
loc, ty, nm, symNm, dataAttr, lenParams, indices);
return alloc;
}

// Let the builder do all the heavy lifting.
Expand Down Expand Up @@ -927,6 +928,19 @@ static void instantiateLocal(Fortran::lower::AbstractConverter &converter,
finalizeAtRuntime(converter, var, symMap);
if (mustBeDefaultInitializedAtRuntime(var))
defaultInitializeAtRuntime(converter, var, symMap);
if (Fortran::semantics::NeedCUDAAlloc(var.getSymbol())) {
auto *builder = &converter.getFirOpBuilder();
mlir::Location loc = converter.getCurrentLocation();
fir::ExtendedValue exv =
converter.getSymbolExtendedValue(var.getSymbol(), &symMap);
auto *sym = &var.getSymbol();
converter.getFctCtx().attachCleanup([builder, loc, exv, sym]() {
cuf::DataAttributeAttr dataAttr =
Fortran::lower::translateSymbolCUFDataAttribute(builder->getContext(),
*sym);
builder->create<cuf::FreeOp>(loc, fir::getBase(exv), dataAttr);
});
}
if (std::optional<VariableCleanUp> cleanup =
needDeallocationOrFinalization(var)) {
auto *builder = &converter.getFirOpBuilder();
Expand All @@ -950,22 +964,10 @@ static void instantiateLocal(Fortran::lower::AbstractConverter &converter,
"trying to deallocate entity not lowered as allocatable");
Fortran::lower::genDeallocateIfAllocated(*converterPtr, *mutableBox,
loc, sym);

});
}
}
if (Fortran::semantics::NeedCUDAAlloc(var.getSymbol())) {
auto *builder = &converter.getFirOpBuilder();
mlir::Location loc = converter.getCurrentLocation();
fir::ExtendedValue exv =
converter.getSymbolExtendedValue(var.getSymbol(), &symMap);
auto *sym = &var.getSymbol();
converter.getFctCtx().attachCleanup([builder, loc, exv, sym]() {
cuf::DataAttributeAttr dataAttr =
Fortran::lower::translateSymbolCUFDataAttribute(builder->getContext(),
*sym);
builder->create<cuf::FreeOp>(loc, fir::getBase(exv), dataAttr);
});
}
}

//===----------------------------------------------------------------===//
Expand Down
6 changes: 4 additions & 2 deletions flang/lib/Optimizer/Dialect/CUF/CUFOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,9 +54,11 @@ template <typename Op>
static llvm::LogicalResult checkCudaAttr(Op op) {
if (op.getDataAttr() == cuf::DataAttribute::Device ||
op.getDataAttr() == cuf::DataAttribute::Managed ||
op.getDataAttr() == cuf::DataAttribute::Unified)
op.getDataAttr() == cuf::DataAttribute::Unified ||
op.getDataAttr() == cuf::DataAttribute::Pinned)
return mlir::success();
return op.emitOpError("expect device, managed or unified cuda attribute");
return op.emitOpError()
<< "expect device, managed, pinned or unified cuda attribute";
}

llvm::LogicalResult cuf::AllocOp::verify() { return checkCudaAttr(*this); }
Expand Down
11 changes: 1 addition & 10 deletions flang/test/Fir/cuf-invalid.fir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -88,18 +88,9 @@ func.func @_QPsub1() {

// -----

func.func @_QPsub1() {
// expected-error@+1{{'cuf.alloc' op expect device, managed or unified cuda attribute}}
%0 = cuf.alloc f32 {bindc_name = "r", data_attr = #cuf.cuda<pinned>, uniq_name = "_QFsub1Er"} -> !fir.ref<f32>
cuf.free %0 : !fir.ref<f32> {data_attr = #cuf.cuda<constant>}
return
}

// -----

func.func @_QPsub1() {
%0 = cuf.alloc f32 {bindc_name = "r", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub1Er"} -> !fir.ref<f32>
// expected-error@+1{{'cuf.free' op expect device, managed or unified cuda attribute}}
// expected-error@+1{{'cuf.free' op expect device, managed, pinned or unified cuda attribute}}
cuf.free %0 : !fir.ref<f32> {data_attr = #cuf.cuda<constant>}
return
}
21 changes: 14 additions & 7 deletions flang/test/Lower/CUDA/cuda-allocatable.cuf
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@ subroutine sub1()
end subroutine

! CHECK-LABEL: func.func @_QPsub1()
! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
! CHECK: %[[BOX:.*]] = cuf.alloc !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub1Ea"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %[[BOX]] {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
! CHECK: fir.call @_FortranAAllocatableSetBounds
! CHECK: %{{.*}} = cuf.allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>} -> i32
Expand All @@ -25,6 +25,7 @@ end subroutine
! CHECK: fir.if %[[NE_C0]] {
! CHECK: %{{.*}} = cuf.deallocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>} -> i32
! CHECK: }
! CHECK: cuf.free %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>}

subroutine sub2()
real, allocatable, managed :: a(:)
Expand All @@ -35,7 +36,7 @@ subroutine sub2()
end subroutine

! CHECK-LABEL: func.func @_QPsub2()
! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub2Ea"}
! CHECK: %[[BOX:.*]] = cuf.alloc !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", data_attr = #cuf.cuda<managed>, uniq_name = "_QFsub2Ea"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %[[BOX]] {data_attr = #cuf.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub2Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
! CHECK: %[[ISTAT:.*]] = fir.alloca i32 {bindc_name = "istat", uniq_name = "_QFsub2Eistat"}
! CHECK: %[[ISTAT_DECL:.*]]:2 = hlfir.declare %[[ISTAT]] {uniq_name = "_QFsub2Eistat"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
Expand All @@ -49,6 +50,7 @@ end subroutine
! CHECK: fir.if %{{.*}} {
! CHECK: %{{.*}} = cuf.deallocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<managed>} -> i32
! CHECK: }
! CHECK: cuf.free %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<managed>}

subroutine sub3()
integer, allocatable, pinned :: a(:,:)
Expand All @@ -57,7 +59,7 @@ subroutine sub3()
end subroutine

! CHECK-LABEL: func.func @_QPsub3()
! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?x?xi32>>> {bindc_name = "a", uniq_name = "_QFsub3Ea"}
! CHECK: %[[BOX:.*]] = cuf.alloc !fir.box<!fir.heap<!fir.array<?x?xi32>>> {bindc_name = "a", data_attr = #cuf.cuda<pinned>, uniq_name = "_QFsub3Ea"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>
! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %[[BOX]] {data_attr = #cuf.cuda<pinned>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub3Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>)
! CHECK: %[[PLOG:.*]] = fir.alloca !fir.logical<4> {bindc_name = "plog", uniq_name = "_QFsub3Eplog"}
! CHECK: %[[PLOG_DECL:.*]]:2 = hlfir.declare %5 {uniq_name = "_QFsub3Eplog"} : (!fir.ref<!fir.logical<4>>) -> (!fir.ref<!fir.logical<4>>, !fir.ref<!fir.logical<4>>)
Expand All @@ -66,6 +68,7 @@ end subroutine
! CHECK: fir.if %{{.*}} {
! CHECK: %{{.*}} = cuf.deallocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>> {data_attr = #cuf.cuda<pinned>} -> i32
! CHECK: }
! CHECK: cuf.free %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>> {data_attr = #cuf.cuda<pinned>}

subroutine sub4()
real, allocatable, device :: a(:)
Expand All @@ -74,7 +77,7 @@ subroutine sub4()
end subroutine

! CHECK-LABEL: func.func @_QPsub4()
! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub4Ea"}
! CHECK: %[[BOX:.*]] = cuf.alloc !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub4Ea"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %0 {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub4Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
! CHECK: %[[ISTREAM:.*]] = fir.alloca i32 {bindc_name = "istream", uniq_name = "_QFsub4Eistream"}
! CHECK: %[[ISTREAM_DECL:.*]]:2 = hlfir.declare %[[ISTREAM]] {uniq_name = "_QFsub4Eistream"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
Expand All @@ -84,6 +87,7 @@ end subroutine
! CHECK: fir.if %{{.*}} {
! CHECK: %{{.*}} = cuf.deallocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>} -> i32
! CHECK: }
! CHECK: cuf.free %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>}

subroutine sub5()
real, allocatable, device :: a(:)
Expand All @@ -92,7 +96,7 @@ subroutine sub5()
end subroutine

! CHECK-LABEL: func.func @_QPsub5()
! CHECK: %[[BOX_A:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub5Ea"}
! CHECK: %[[BOX_A:.*]] = cuf.alloc !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub5Ea"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
! CHECK: %[[BOX_A_DECL:.*]]:2 = hlfir.declare %[[BOX]] {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub5Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
! CHECK: %[[BOX_B:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "b", uniq_name = "_QFsub5Eb"}
! CHECK: %[[BOX_B_DECL:.*]]:2 = hlfir.declare %[[BOX_B]] {fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub5Eb"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
Expand All @@ -104,6 +108,7 @@ end subroutine
! CHECK: fir.if %{{.*}} {
! CHECK: %{{.*}} = cuf.deallocate %[[BOX_A_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>} -> i32
! CHECK: }
! CHECK: cuf.free %[[BOX_A_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>}

subroutine sub6()
real, allocatable, device :: a(:)
Expand All @@ -112,7 +117,7 @@ subroutine sub6()
end subroutine

! CHECK-LABEL: func.func @_QPsub6()
! CHECK: %[[BOX_A:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub6Ea"}
! CHECK: %[[BOX_A:.*]] = cuf.alloc !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub6Ea"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
! CHECK: %[[BOX_A_DECL:.*]]:2 = hlfir.declare %[[BOX]] {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub6Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
! CHECK: %[[BOX_B:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "b", uniq_name = "_QFsub6Eb"}
! CHECK: %[[BOX_B_DECL:.*]]:2 = hlfir.declare %[[BOX_B]] {fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub6Eb"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
Expand All @@ -122,6 +127,7 @@ end subroutine
! CHECK: fir.if %{{.*}} {
! CHECK: %{{.*}} = cuf.deallocate %[[BOX_A_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>} -> i32
! CHECK: }
! CHECK: cuf.free %[[BOX_A_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>}

subroutine sub7()
real, allocatable, device :: a(:)
Expand All @@ -133,7 +139,7 @@ subroutine sub7()
end subroutine

! CHECK-LABEL: func.func @_QPsub7()
! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub7Ea"}
! CHECK: %[[BOX:.*]] = cuf.alloc !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", data_attr = #cuf.cuda<device>, uniq_name = "_QFsub7Ea"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %[[BOX]] {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub7Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
! CHECK: %[[ERR:.*]] = fir.alloca !fir.char<1,50> {bindc_name = "err", uniq_name = "_QFsub7Eerr"}
! CHECK: %[[ERR_DECL:.*]]:2 = hlfir.declare %[[ERR]] typeparams %{{.*}} {uniq_name = "_QFsub7Eerr"} : (!fir.ref<!fir.char<1,50>>, index) -> (!fir.ref<!fir.char<1,50>>, !fir.ref<!fir.char<1,50>>)
Expand All @@ -150,3 +156,4 @@ end subroutine
! CHECK: fir.if %{{.*}} {
! CHECK: %{{.*}} = cuf.deallocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>} -> i32
! CHECK: }
! CHECK: cuf.free %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {data_attr = #cuf.cuda<device>}
Loading