Skip to content

[flang][OpenMP] use reduction alloc region #102525

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 4 commits into from
Aug 22, 2024
Merged

[flang][OpenMP] use reduction alloc region #102525

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
57e4735
[mlir][OpenMP] Add optional alloc region to reduction decl
tblah Jul 29, 2024
b916864
Fix documentation
tblah Aug 22, 2024
9e37da3
[mlir][OpenMP] Convert reduction alloc region to LLVMIR
tblah Jul 31, 2024
ee4a8e0
[flang][OpenMP] use reduction alloc region
tblah Jul 30, 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
86 changes: 63 additions & 23 deletions flang/lib/Lower/OpenMP/ReductionProcessor.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -489,23 +489,57 @@ static mlir::Type unwrapSeqOrBoxedType(mlir::Type ty) {
return ty;
}

static mlir::Value
createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
mlir::omp::DeclareReductionOp &reductionDecl,
const ReductionProcessor::ReductionIdentifier redId,
mlir::Type type, bool isByRef) {
static void createReductionAllocAndInitRegions(
fir::FirOpBuilder &builder, mlir::Location loc,
mlir::omp::DeclareReductionOp &reductionDecl,
const ReductionProcessor::ReductionIdentifier redId, mlir::Type type,
bool isByRef) {
auto yield = [&](mlir::Value ret) {
builder.create<mlir::omp::YieldOp>(loc, ret);
};

mlir::Block *allocBlock = nullptr;
mlir::Block *initBlock = nullptr;
if (isByRef) {
allocBlock =
builder.createBlock(&reductionDecl.getAllocRegion(),
reductionDecl.getAllocRegion().end(), {}, {});
initBlock = builder.createBlock(&reductionDecl.getInitializerRegion(),
reductionDecl.getInitializerRegion().end(),
{type, type}, {loc, loc});
} else {
initBlock = builder.createBlock(&reductionDecl.getInitializerRegion(),
reductionDecl.getInitializerRegion().end(),
{type}, {loc});
}

mlir::Type ty = fir::unwrapRefType(type);
builder.setInsertionPointToEnd(initBlock);
mlir::Value initValue = ReductionProcessor::getReductionInitValue(
loc, unwrapSeqOrBoxedType(ty), redId, builder);

if (fir::isa_trivial(ty)) {
if (isByRef) {
mlir::Value alloca = builder.create<fir::AllocaOp>(loc, ty);
builder.createStoreWithConvert(loc, initValue, alloca);
return alloca;
// alloc region
{
builder.setInsertionPointToEnd(allocBlock);
mlir::Value alloca = builder.create<fir::AllocaOp>(loc, ty);
yield(alloca);
}

// init region
{
builder.setInsertionPointToEnd(initBlock);
// block arg is mapped to the alloca yielded from the alloc region
mlir::Value alloc = reductionDecl.getInitializerAllocArg();
builder.createStoreWithConvert(loc, initValue, alloc);
yield(alloc);
}
return;
}
// by val
return initValue;
yield(initValue);
return;
}

// check if an allocatable box is unallocated. If so, initialize the boxAlloca
Expand All @@ -520,10 +554,10 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
// fir.store %something to %box_alloca
// }
// omp.yield %box_alloca
mlir::Value blockArg =
builder.loadIfRef(loc, builder.getBlock()->getArgument(0));
mlir::Value moldArg =
builder.loadIfRef(loc, reductionDecl.getInitializerMoldArg());
auto handleNullAllocatable = [&](mlir::Value boxAlloca) -> fir::IfOp {
mlir::Value addr = builder.create<fir::BoxAddrOp>(loc, blockArg);
mlir::Value addr = builder.create<fir::BoxAddrOp>(loc, moldArg);
mlir::Value isNotAllocated = builder.genIsNullAddr(loc, addr);
fir::IfOp ifOp = builder.create<fir::IfOp>(loc, isNotAllocated,
/*withElseRegion=*/true);
Expand All @@ -539,7 +573,17 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
assert(isByRef && "passing boxes by value is unsupported");
bool isAllocatableOrPointer =
mlir::isa<fir::HeapType, fir::PointerType>(boxTy.getEleTy());
mlir::Value boxAlloca = builder.create<fir::AllocaOp>(loc, ty);

// alloc region
{
builder.setInsertionPointToEnd(allocBlock);
mlir::Value boxAlloca = builder.create<fir::AllocaOp>(loc, ty);
yield(boxAlloca);
}

// init region
builder.setInsertionPointToEnd(initBlock);
mlir::Value boxAlloca = reductionDecl.getInitializerAllocArg();
mlir::Type innerTy = fir::unwrapRefType(boxTy.getEleTy());
if (fir::isa_trivial(innerTy)) {
// boxed non-sequence value e.g. !fir.box<!fir.heap<i32>>
Expand All @@ -558,7 +602,8 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
createReductionCleanupRegion(builder, loc, reductionDecl);
builder.restoreInsertionPoint(insPt);
builder.setInsertionPointAfter(ifUnallocated);
return boxAlloca;
yield(boxAlloca);
return;
}
innerTy = fir::extractSequenceType(boxTy);
if (!mlir::isa<fir::SequenceType>(innerTy))
Expand All @@ -571,7 +616,7 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
}

// Create the private copy from the initial fir.box:
mlir::Value loadedBox = builder.loadIfRef(loc, blockArg);
mlir::Value loadedBox = builder.loadIfRef(loc, moldArg);
hlfir::Entity source = hlfir::Entity{loadedBox};

// Allocating on the heap in case the whole reduction is nested inside of a
Expand Down Expand Up @@ -616,7 +661,8 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
builder.create<fir::StoreOp>(loc, box, boxAlloca);
if (ifUnallocated)
builder.setInsertionPointAfter(ifUnallocated);
return boxAlloca;
yield(boxAlloca);
return;
}

TODO(loc, "createReductionInitRegion for unsupported type");
Expand All @@ -643,13 +689,7 @@ mlir::omp::DeclareReductionOp ReductionProcessor::createDeclareReduction(

decl = modBuilder.create<mlir::omp::DeclareReductionOp>(loc, reductionOpName,
type);
builder.createBlock(&decl.getInitializerRegion(),
decl.getInitializerRegion().end(), {type}, {loc});
builder.setInsertionPointToEnd(&decl.getInitializerRegion().back());

mlir::Value init =
createReductionInitRegion(builder, loc, decl, redId, type, isByRef);
builder.create<mlir::omp::YieldOp>(loc, init);
createReductionAllocAndInitRegions(builder, loc, decl, redId, type, isByRef);

builder.createBlock(&decl.getReductionRegion(),
decl.getReductionRegion().end(), {type, type},
Expand Down
2 changes: 1 addition & 1 deletion flang/test/Lower/OpenMP/delayed-privatization-reduction-byref.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@ subroutine red_and_delayed_private
! CHECK-SAME: @[[PRIVATIZER_SYM:.*]] : !fir.ref<i32> alloc {

! CHECK-LABEL: omp.declare_reduction
! CHECK-SAME: @[[REDUCTION_SYM:.*]] : !fir.ref<i32> init
! CHECK-SAME: @[[REDUCTION_SYM:.*]] : !fir.ref<i32> alloc

! CHECK-LABEL: _QPred_and_delayed_private
! CHECK: omp.parallel
Expand Down
20 changes: 12 additions & 8 deletions flang/test/Lower/OpenMP/parallel-reduction-add-byref.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,12 +3,14 @@

!CHECK-LABEL: omp.declare_reduction
!CHECK-SAME: @[[RED_F32_NAME:.*]] : !fir.ref<f32>
!CHECK-SAME: init {
!CHECK: ^bb0(%{{.*}}: !fir.ref<f32>):
!CHECK: %[[C0_1:.*]] = arith.constant 0.000000e+00 : f32
!CHECK-SAME: alloc {
!CHECK: %[[REF:.*]] = fir.alloca f32
!CHECKL fir.store [[%C0_1]] to %[[REF]] : !fir.ref<f32>
!CHECK: omp.yield(%[[REF]] : !fir.ref<f32>)
!CHECK-LABEL: } init {
!CHECK: ^bb0(%{{.*}}: !fir.ref<f32>, %[[ALLOC:.*]]: !fir.ref<f32>):
!CHECK: %[[C0_1:.*]] = arith.constant 0.000000e+00 : f32
!CHECKL fir.store [[%C0_1]] to %[[ALLOC]] : !fir.ref<f32>
!CHECK: omp.yield(%[[ALLOC]] : !fir.ref<f32>)
!CHECK: } combiner {
!CHECK: ^bb0(%[[ARG0:.*]]: !fir.ref<f32>, %[[ARG1:.*]]: !fir.ref<f32>):
!CHECK: %[[LD0:.*]] = fir.load %[[ARG0]] : !fir.ref<f32>
Expand All @@ -20,12 +22,14 @@

!CHECK-LABEL: omp.declare_reduction
!CHECK-SAME: @[[RED_I32_NAME:.*]] : !fir.ref<i32>
!CHECK-SAME: init {
!CHECK: ^bb0(%{{.*}}: !fir.ref<i32>):
!CHECK: %[[C0_1:.*]] = arith.constant 0 : i32
!CHECK-SAME: alloc {
!CHECK: %[[REF:.*]] = fir.alloca i32
!CHECK: fir.store %[[C0_1]] to %[[REF]] : !fir.ref<i32>
!CHECK: omp.yield(%[[REF]] : !fir.ref<i32>)
!CHECK-LABEL: } init {
!CHECK: ^bb0(%{{.*}}: !fir.ref<i32>, %[[ALLOC:.*]]: !fir.ref<i32>):
!CHECK: %[[C0_1:.*]] = arith.constant 0 : i32
!CHECK: fir.store %[[C0_1]] to %[[ALLOC]] : !fir.ref<i32>
!CHECK: omp.yield(%[[ALLOC]] : !fir.ref<i32>)
!CHECK: } combiner {
!CHECK: ^bb0(%[[ARG0:.*]]: !fir.ref<i32>, %[[ARG1:.*]]: !fir.ref<i32>):
!CHECK: %[[LD0:.*]] = fir.load %[[ARG0]] : !fir.ref<i32>
Expand Down
14 changes: 8 additions & 6 deletions flang/test/Lower/OpenMP/parallel-reduction-allocatable-array.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,18 +18,20 @@ program reduce

end program

! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_heap_Uxi32 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>> init {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>):
! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_heap_Uxi32 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>> alloc {
! CHECK: %[[VAL_10:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xi32>>>
! CHECK: omp.yield(%[[VAL_10]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>)
! CHECK-LABEL: } init {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>, %[[ALLOC:.*]]: !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>):
! CHECK: %[[VAL_1:.*]] = arith.constant 0 : i32
! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
! CHECK: %[[VAL_10:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xi32>>>
! CHECK: %[[ADDR:.*]] = fir.box_addr %[[VAL_2]] : (!fir.box<!fir.heap<!fir.array<?xi32>>>) -> !fir.heap<!fir.array<?xi32>>
! CHECK: %[[ADDRI:.*]] = fir.convert %[[ADDR]] : (!fir.heap<!fir.array<?xi32>>) -> i64
! CHECK: %[[C0_I64:.*]] = arith.constant 0 : i64
! CHECK: %[[IS_NULL:.*]] = arith.cmpi eq, %[[ADDRI]], %[[C0_I64]] : i64
! CHECK: fir.if %[[IS_NULL]] {
! CHECK: %[[NULL_BOX:.*]] = fir.embox %[[ADDR]] : (!fir.heap<!fir.array<?xi32>>) -> !fir.box<!fir.heap<!fir.array<?xi32>>>
! CHECK: fir.store %[[NULL_BOX]] to %[[VAL_10]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
! CHECK: fir.store %[[NULL_BOX]] to %[[ALLOC]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
! CHECK: } else {
! CHECK: %[[VAL_3:.*]] = arith.constant 0 : index
! CHECK: %[[VAL_4:.*]]:3 = fir.box_dims %[[VAL_2]], %[[VAL_3]] : (!fir.box<!fir.heap<!fir.array<?xi32>>>, index) -> (index, index, index)
Expand All @@ -42,9 +44,9 @@ program reduce
! CHECK: %[[SHIFT:.*]] = fir.shape_shift %[[DIMS]]#0, %[[DIMS]]#1 : (index, index) -> !fir.shapeshift<1>
! CHECK: %[[REBOX:.*]] = fir.rebox %[[VAL_8]]#0(%[[SHIFT]]) : (!fir.box<!fir.array<?xi32>>, !fir.shapeshift<1>) -> !fir.box<!fir.heap<!fir.array<?xi32>>>
! CHECK: hlfir.assign %[[VAL_1]] to %[[REBOX]] : i32, !fir.box<!fir.heap<!fir.array<?xi32>>>
! CHECK: fir.store %[[REBOX]] to %[[VAL_10]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
! CHECK: fir.store %[[REBOX]] to %[[ALLOC]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
! CHECK: }
! CHECK: omp.yield(%[[VAL_10]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>)
! CHECK: omp.yield(%[[ALLOC]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>)
! CHECK: } combiner {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>, %[[VAL_1:.*]]: !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>):
! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
Expand Down
12 changes: 7 additions & 5 deletions flang/test/Lower/OpenMP/parallel-reduction-array-lb.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,11 +12,13 @@ program reduce

end program

! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_3x2xi32 : !fir.ref<!fir.box<!fir.array<3x2xi32>>> init {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3x2xi32>>>):
! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_3x2xi32 : !fir.ref<!fir.box<!fir.array<3x2xi32>>> alloc {
! CHECK: %[[VAL_15:.*]] = fir.alloca !fir.box<!fir.array<3x2xi32>>
! CHECK: omp.yield(%[[VAL_15]] : !fir.ref<!fir.box<!fir.array<3x2xi32>>>)
! CHECK-LABEL: } init {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3x2xi32>>>, %[[ALLOC:.*]]: !fir.ref<!fir.box<!fir.array<3x2xi32>>>):
! CHECK: %[[VAL_1:.*]] = arith.constant 0 : i32
! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3x2xi32>>>
! CHECK: %[[VAL_15:.*]] = fir.alloca !fir.box<!fir.array<3x2xi32>>
! CHECK: %[[VAL_3:.*]] = arith.constant 3 : index
! CHECK: %[[VAL_4:.*]] = arith.constant 2 : index
! CHECK: %[[VAL_5:.*]] = fir.shape %[[VAL_3]], %[[VAL_4]] : (index, index) -> !fir.shape<2>
Expand All @@ -30,8 +32,8 @@ program reduce
! CHECK: %[[VAL_13:.*]] = fir.shape_shift %[[VAL_10]]#0, %[[VAL_10]]#1, %[[VAL_12]]#0, %[[VAL_12]]#1 : (index, index, index, index) -> !fir.shapeshift<2>
! CHECK: %[[VAL_14:.*]] = fir.embox %[[VAL_8]]#0(%[[VAL_13]]) : (!fir.heap<!fir.array<3x2xi32>>, !fir.shapeshift<2>) -> !fir.box<!fir.array<3x2xi32>>
! CHECK: hlfir.assign %[[VAL_1]] to %[[VAL_14]] : i32, !fir.box<!fir.array<3x2xi32>>
! CHECK: fir.store %[[VAL_14]] to %[[VAL_15]] : !fir.ref<!fir.box<!fir.array<3x2xi32>>>
! CHECK: omp.yield(%[[VAL_15]] : !fir.ref<!fir.box<!fir.array<3x2xi32>>>)
! CHECK: fir.store %[[VAL_14]] to %[[ALLOC]] : !fir.ref<!fir.box<!fir.array<3x2xi32>>>
! CHECK: omp.yield(%[[ALLOC]] : !fir.ref<!fir.box<!fir.array<3x2xi32>>>)
! CHECK: } combiner {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3x2xi32>>>, %[[VAL_1:.*]]: !fir.ref<!fir.box<!fir.array<3x2xi32>>>):
! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3x2xi32>>>
Expand Down
12 changes: 7 additions & 5 deletions flang/test/Lower/OpenMP/parallel-reduction-array.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,11 +13,13 @@ program reduce
print *,i
end program

! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_3xi32 : !fir.ref<!fir.box<!fir.array<3xi32>>> init {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_3xi32 : !fir.ref<!fir.box<!fir.array<3xi32>>> alloc {
! CHECK: %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<3xi32>>
! CHECK: omp.yield(%[[VAL_8]] : !fir.ref<!fir.box<!fir.array<3xi32>>>)
! CHECK-LABEL: } init {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>, %[[ALLOC:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
! CHECK: %[[VAL_2:.*]] = arith.constant 0 : i32
! CHECK: %[[VAL_3:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
! CHECK: %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<3xi32>>
! CHECK: %[[VAL_4:.*]] = arith.constant 3 : index
! CHECK: %[[VAL_5:.*]] = fir.shape %[[VAL_4]] : (index) -> !fir.shape<1>
! CHECK: %[[VAL_1:.*]] = fir.allocmem !fir.array<3xi32> {bindc_name = ".tmp", uniq_name = ""}
Expand All @@ -29,8 +31,8 @@ program reduce
! CHECK: %[[SHIFT:.*]] = fir.shape_shift %[[DIMS]]#0, %[[DIMS]]#1 : (index, index) -> !fir.shapeshift<1>
! CHECK: %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[SHIFT]]) : (!fir.heap<!fir.array<3xi32>>, !fir.shapeshift<1>) -> !fir.box<!fir.array<3xi32>>
! CHECK: hlfir.assign %[[VAL_2]] to %[[VAL_7]] : i32, !fir.box<!fir.array<3xi32>>
! CHECK: fir.store %[[VAL_7]] to %[[VAL_8]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
! CHECK: omp.yield(%[[VAL_8]] : !fir.ref<!fir.box<!fir.array<3xi32>>>)
! CHECK: fir.store %[[VAL_7]] to %[[ALLOC]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
! CHECK: omp.yield(%[[ALLOC]] : !fir.ref<!fir.box<!fir.array<3xi32>>>)
! CHECK: } combiner {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>, %[[VAL_1:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
Expand Down
12 changes: 7 additions & 5 deletions flang/test/Lower/OpenMP/parallel-reduction-array2.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,11 +13,13 @@ program reduce
print *,i
end program

! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_3xi32 : !fir.ref<!fir.box<!fir.array<3xi32>>> init {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_3xi32 : !fir.ref<!fir.box<!fir.array<3xi32>>> alloc {
! CHECK: %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<3xi32>>
! CHECK: omp.yield(%[[VAL_8]] : !fir.ref<!fir.box<!fir.array<3xi32>>>)
! CHECK-LABEL: } init {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>, %[[ALLOC:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
! CHECK: %[[VAL_2:.*]] = arith.constant 0 : i32
! CHECK: %[[VAL_3:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
! CHECK: %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<3xi32>>
! CHECK: %[[VAL_4:.*]] = arith.constant 3 : index
! CHECK: %[[VAL_5:.*]] = fir.shape %[[VAL_4]] : (index) -> !fir.shape<1>
! CHECK: %[[VAL_1:.*]] = fir.allocmem !fir.array<3xi32>
Expand All @@ -28,8 +30,8 @@ program reduce
! CHECK: %[[SHIFT:.*]] = fir.shape_shift %[[DIMS]]#0, %[[DIMS]]#1 : (index, index) -> !fir.shapeshift<1>
! CHECK: %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[SHIFT]]) : (!fir.heap<!fir.array<3xi32>>, !fir.shapeshift<1>) -> !fir.box<!fir.array<3xi32>>
! CHECK: hlfir.assign %[[VAL_2]] to %[[VAL_7]] : i32, !fir.box<!fir.array<3xi32>>
! CHECK: fir.store %[[VAL_7]] to %[[VAL_8]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
! CHECK: omp.yield(%[[VAL_8]] : !fir.ref<!fir.box<!fir.array<3xi32>>>)
! CHECK: fir.store %[[VAL_7]] to %[[ALLOC]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
! CHECK: omp.yield(%[[ALLOC]] : !fir.ref<!fir.box<!fir.array<3xi32>>>)
! CHECK: } combiner {
! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>, %[[VAL_1:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
Expand Down
12 changes: 7 additions & 5 deletions flang/test/Lower/OpenMP/parallel-reduction-byref.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,12 +2,14 @@
! RUN: %flang_fc1 -emit-hlfir -fopenmp -mmlir --force-byref-reduction -o - %s 2>&1 | FileCheck %s

!CHECK: omp.declare_reduction @[[REDUCTION_DECLARE:[_a-z0-9]+]] : !fir.ref<i32>
!CHECK-SAME: init {
!CHECK: ^bb0(%{{.*}}: !fir.ref<i32>):
!CHECK: %[[I0:[_a-z0-9]+]] = arith.constant 0 : i32
!CHECK: %[[REF:.*]] = fir.alloca i32
!CHECKL fir.store [[%I0]] to %[[REF]] : !fir.ref<i32>
!CHECK-SAME: alloc {
!CHECK: %[[REF:.*]] = fir.alloca i32
!CHECK: omp.yield(%[[REF]] : !fir.ref<i32>)
!CHECK-LABEL: } init {
!CHECK: ^bb0(%{{.*}}: !fir.ref<i32>, %[[ALLOC:.*]]: !fir.ref<i32>):
!CHECK: %[[I0:[_a-z0-9]+]] = arith.constant 0 : i32
!CHECKL fir.store [[%I0]] to %[[ALLOC]] : !fir.ref<i32>
!CHECK: omp.yield(%[[ALLOC]] : !fir.ref<i32>)
!CHECK: } combiner {
!CHECK: ^bb0(%[[C0:[_a-z0-9]+]]: !fir.ref<i32>, %[[C1:[_a-z0-9]+]]: !fir.ref<i32>):
!CHECK: %[[LD0:.*]] = fir.load %[[C0]] : !fir.ref<i32>
Expand Down
4 changes: 2 additions & 2 deletions flang/test/Lower/OpenMP/parallel-reduction-mixed.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,11 +24,11 @@ end subroutine proc
!CHECK: %[[TID_LOCAL:.*]] = alloca i32
!CHECK: %[[TID:.*]] = load i32, ptr %[[TID_ADDR]]
!CHECK: store i32 %[[TID]], ptr %[[TID_LOCAL]]
!CHECK: %[[I_priv:.*]] = alloca i32
!CHECK: %[[F_priv:.*]] = alloca ptr
!CHECK: %[[I_priv:.*]] = alloca i32
!CHECK: store ptr %{{.*}}, ptr %[[F_priv]]

!CHECK: omp.reduction.init:
!CHECK: store ptr %{{.*}}, ptr %[[F_priv]]
!CHECK: store i32 0, ptr %[[I_priv]]

!CHECK: omp.par.region:
Expand Down
Loading
Loading