[flang-commits] [flang] [flang][OpenMP] Support reduction of POINTER variables (PR #95148)
Tom Eccles via flang-commits
flang-commits at lists.llvm.org
Thu Jun 13 06:17:35 PDT 2024
https://github.com/tblah updated https://github.com/llvm/llvm-project/pull/95148
>From f356f64ee0c34611a396ebd4df03d8f7453bde34 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles at arm.com>
Date: Tue, 11 Jun 2024 16:58:39 +0000
Subject: [PATCH 1/5] [flang][OpenMP] Support reduction of POINTER variables
Just treat them the same as ALLOCATABLE. gfortran doesn't
allow POINTER objects in a REDUCTION clause, but so far as I can tell
the standard explicitly allows it (openmp5.2 section 5.5.5).
---
flang/lib/Lower/OpenMP/ReductionProcessor.cpp | 22 +--
.../parallel-reduction-pointer-array.f90 | 131 ++++++++++++++++++
.../Lower/OpenMP/wsloop-reduction-pointer.f90 | 110 +++++++++++++++
3 files changed, 255 insertions(+), 8 deletions(-)
create mode 100644 flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
create mode 100644 flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
diff --git a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
index 60e933f5bc1f7..b32036f4b5f4a 100644
--- a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
@@ -332,7 +332,9 @@ static void genBoxCombiner(fir::FirOpBuilder &builder, mlir::Location loc,
fir::unwrapRefType(boxTy.getEleTy()));
fir::HeapType heapTy =
mlir::dyn_cast_or_null<fir::HeapType>(boxTy.getEleTy());
- if ((!seqTy || seqTy.hasUnknownShape()) && !heapTy)
+ fir::PointerType ptrTy =
+ mlir::dyn_cast_or_null<fir::PointerType>(boxTy.getEleTy());
+ if ((!seqTy || seqTy.hasUnknownShape()) && !heapTy && !ptrTy)
TODO(loc, "Unsupported boxed type in OpenMP reduction");
// load fir.ref<fir.box<...>>
@@ -340,7 +342,7 @@ static void genBoxCombiner(fir::FirOpBuilder &builder, mlir::Location loc,
lhs = builder.create<fir::LoadOp>(loc, lhs);
rhs = builder.create<fir::LoadOp>(loc, rhs);
- if (heapTy && !seqTy) {
+ if ((heapTy || ptrTy) && !seqTy) {
// get box contents (heap pointers)
lhs = builder.create<fir::BoxAddrOp>(loc, lhs);
rhs = builder.create<fir::BoxAddrOp>(loc, rhs);
@@ -350,8 +352,10 @@ static void genBoxCombiner(fir::FirOpBuilder &builder, mlir::Location loc,
lhs = builder.create<fir::LoadOp>(loc, lhs);
rhs = builder.create<fir::LoadOp>(loc, rhs);
+ mlir::Type eleTy = heapTy ? heapTy.getEleTy() : ptrTy.getEleTy();
+
mlir::Value result = ReductionProcessor::createScalarCombiner(
- builder, loc, redId, heapTy.getEleTy(), lhs, rhs);
+ builder, loc, redId, eleTy, lhs, rhs);
builder.create<fir::StoreOp>(loc, result, lhsValAddr);
builder.create<mlir::omp::YieldOp>(loc, lhsAddr);
return;
@@ -439,7 +443,7 @@ createReductionCleanupRegion(fir::FirOpBuilder &builder, mlir::Location loc,
mlir::Type valTy = fir::unwrapRefType(redTy);
if (auto boxTy = mlir::dyn_cast_or_null<fir::BaseBoxType>(valTy)) {
- if (!mlir::isa<fir::HeapType>(boxTy.getEleTy())) {
+ if (!mlir::isa<fir::HeapType, fir::PointerType>(boxTy.getEleTy())) {
mlir::Type innerTy = fir::extractSequenceType(boxTy);
if (!mlir::isa<fir::SequenceType>(innerTy))
typeError();
@@ -533,12 +537,13 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
// all arrays are boxed
if (auto boxTy = mlir::dyn_cast_or_null<fir::BaseBoxType>(ty)) {
assert(isByRef && "passing boxes by value is unsupported");
- bool isAllocatable = mlir::isa<fir::HeapType>(boxTy.getEleTy());
+ bool isAllocatableOrPointer =
+ mlir::isa<fir::HeapType, fir::PointerType>(boxTy.getEleTy());
mlir::Value boxAlloca = builder.create<fir::AllocaOp>(loc, ty);
mlir::Type innerTy = fir::unwrapRefType(boxTy.getEleTy());
if (fir::isa_trivial(innerTy)) {
// boxed non-sequence value e.g. !fir.box<!fir.heap<i32>>
- if (!isAllocatable)
+ if (!isAllocatableOrPointer)
TODO(loc, "Reduction of non-allocatable trivial typed box");
fir::IfOp ifUnallocated = handleNullAllocatable(boxAlloca);
@@ -560,7 +565,7 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
TODO(loc, "Unsupported boxed type for reduction");
fir::IfOp ifUnallocated{nullptr};
- if (isAllocatable) {
+ if (isAllocatableOrPointer) {
ifUnallocated = handleNullAllocatable(boxAlloca);
builder.setInsertionPointToStart(&ifUnallocated.getElseRegion().front());
}
@@ -587,7 +592,8 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
mlir::OpBuilder::InsertionGuard guard(builder);
createReductionCleanupRegion(builder, loc, reductionDecl);
} else {
- assert(!isAllocatable && "Allocatable arrays must be heap allocated");
+ assert(!isAllocatableOrPointer &&
+ "Allocatable arrays must be heap allocated");
}
// Put the temporary inside of a box:
diff --git a/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90 b/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
new file mode 100644
index 0000000000000..dfdc50d0c116d
--- /dev/null
+++ b/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
@@ -0,0 +1,131 @@
+! RUN: bbc -emit-hlfir -fopenmp -o - %s | FileCheck %s
+! RUN: %flang_fc1 -emit-hlfir -fopenmp -o - %s | FileCheck %s
+
+program reduce
+integer :: i = 0
+integer, dimension(:), pointer :: r
+
+allocate(r(2))
+
+!$omp parallel do reduction(+:r)
+do i=0,10
+ r(1) = i
+ r(2) = -i
+enddo
+!$omp end parallel do
+
+print *,r
+deallocate(r)
+
+end program
+
+! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_ptr_Uxi32 : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>> init {
+! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>):
+! CHECK: %[[VAL_1:.*]] = arith.constant 0 : i32
+! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: %[[VAL_3:.*]] = fir.alloca !fir.box<!fir.ptr<!fir.array<?xi32>>>
+! CHECK: %[[VAL_4:.*]] = fir.box_addr %[[VAL_2]] : (!fir.box<!fir.ptr<!fir.array<?xi32>>>) -> !fir.ptr<!fir.array<?xi32>>
+! CHECK: %[[VAL_5:.*]] = fir.convert %[[VAL_4]] : (!fir.ptr<!fir.array<?xi32>>) -> i64
+! CHECK: %[[VAL_6:.*]] = arith.constant 0 : i64
+! CHECK: %[[VAL_7:.*]] = arith.cmpi eq, %[[VAL_5]], %[[VAL_6]] : i64
+! CHECK: fir.if %[[VAL_7]] {
+! CHECK: %[[VAL_8:.*]] = fir.embox %[[VAL_4]] : (!fir.ptr<!fir.array<?xi32>>) -> !fir.box<!fir.ptr<!fir.array<?xi32>>>
+! CHECK: fir.store %[[VAL_8]] to %[[VAL_3]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: } else {
+! CHECK: %[[VAL_9:.*]] = arith.constant 0 : index
+! CHECK: %[[VAL_10:.*]]:3 = fir.box_dims %[[VAL_2]], %[[VAL_9]] : (!fir.box<!fir.ptr<!fir.array<?xi32>>>, index) -> (index, index, index)
+! CHECK: %[[VAL_11:.*]] = fir.shape %[[VAL_10]]#1 : (index) -> !fir.shape<1>
+! CHECK: %[[VAL_12:.*]] = fir.allocmem !fir.array<?xi32>, %[[VAL_10]]#1 {bindc_name = ".tmp", uniq_name = ""}
+! CHECK: %[[VAL_13:.*]] = arith.constant true
+! CHECK: %[[VAL_14:.*]]:2 = hlfir.declare %[[VAL_12]](%[[VAL_11]]) {uniq_name = ".tmp"} : (!fir.heap<!fir.array<?xi32>>, !fir.shape<1>) -> (!fir.box<!fir.array<?xi32>>, !fir.heap<!fir.array<?xi32>>)
+! CHECK: %[[VAL_15:.*]] = arith.constant 0 : index
+! CHECK: %[[VAL_16:.*]]:3 = fir.box_dims %[[VAL_2]], %[[VAL_15]] : (!fir.box<!fir.ptr<!fir.array<?xi32>>>, index) -> (index, index, index)
+! CHECK: %[[VAL_17:.*]] = fir.shape_shift %[[VAL_16]]#0, %[[VAL_16]]#1 : (index, index) -> !fir.shapeshift<1>
+! CHECK: %[[VAL_18:.*]] = fir.rebox %[[VAL_14]]#0(%[[VAL_17]]) : (!fir.box<!fir.array<?xi32>>, !fir.shapeshift<1>) -> !fir.box<!fir.ptr<!fir.array<?xi32>>>
+! CHECK: hlfir.assign %[[VAL_1]] to %[[VAL_18]] : i32, !fir.box<!fir.ptr<!fir.array<?xi32>>>
+! CHECK: fir.store %[[VAL_18]] to %[[VAL_3]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: }
+! CHECK: omp.yield(%[[VAL_3]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>)
+! CHECK-LABEL: } combiner {
+! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, %[[VAL_1:.*]]: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>):
+! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: %[[VAL_3:.*]] = fir.load %[[VAL_1]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: %[[VAL_4:.*]] = arith.constant 0 : index
+! CHECK: %[[VAL_5:.*]]:3 = fir.box_dims %[[VAL_2]], %[[VAL_4]] : (!fir.box<!fir.ptr<!fir.array<?xi32>>>, index) -> (index, index, index)
+! CHECK: %[[VAL_6:.*]] = fir.shape_shift %[[VAL_5]]#0, %[[VAL_5]]#1 : (index, index) -> !fir.shapeshift<1>
+! CHECK: %[[VAL_7:.*]] = arith.constant 1 : index
+! CHECK: fir.do_loop %[[VAL_8:.*]] = %[[VAL_7]] to %[[VAL_5]]#1 step %[[VAL_7]] unordered {
+! CHECK: %[[VAL_9:.*]] = fir.array_coor %[[VAL_2]](%[[VAL_6]]) %[[VAL_8]] : (!fir.box<!fir.ptr<!fir.array<?xi32>>>, !fir.shapeshift<1>, index) -> !fir.ref<i32>
+! CHECK: %[[VAL_10:.*]] = fir.array_coor %[[VAL_3]](%[[VAL_6]]) %[[VAL_8]] : (!fir.box<!fir.ptr<!fir.array<?xi32>>>, !fir.shapeshift<1>, index) -> !fir.ref<i32>
+! CHECK: %[[VAL_11:.*]] = fir.load %[[VAL_9]] : !fir.ref<i32>
+! CHECK: %[[VAL_12:.*]] = fir.load %[[VAL_10]] : !fir.ref<i32>
+! CHECK: %[[VAL_13:.*]] = arith.addi %[[VAL_11]], %[[VAL_12]] : i32
+! CHECK: fir.store %[[VAL_13]] to %[[VAL_9]] : !fir.ref<i32>
+! CHECK: }
+! CHECK: omp.yield(%[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>)
+! CHECK-LABEL: } cleanup {
+! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>):
+! CHECK: %[[VAL_1:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: %[[VAL_2:.*]] = fir.box_addr %[[VAL_1]] : (!fir.box<!fir.ptr<!fir.array<?xi32>>>) -> !fir.ptr<!fir.array<?xi32>>
+! CHECK: %[[VAL_3:.*]] = fir.convert %[[VAL_2]] : (!fir.ptr<!fir.array<?xi32>>) -> i64
+! CHECK: %[[VAL_4:.*]] = arith.constant 0 : i64
+! CHECK: %[[VAL_5:.*]] = arith.cmpi ne, %[[VAL_3]], %[[VAL_4]] : i64
+! CHECK: fir.if %[[VAL_5]] {
+! CHECK: %[[VAL_6:.*]] = fir.convert %[[VAL_2]] : (!fir.ptr<!fir.array<?xi32>>) -> !fir.heap<!fir.array<?xi32>>
+! CHECK: fir.freemem %[[VAL_6]] : !fir.heap<!fir.array<?xi32>>
+! CHECK: }
+! CHECK: omp.yield
+! CHECK: }
+
+! CHECK-LABEL: func.func @_QQmain() attributes {fir.bindc_name = "reduce"} {
+! CHECK: %[[VAL_0:.*]] = fir.address_of(@_QFEi) : !fir.ref<i32>
+! CHECK: %[[VAL_1:.*]]:2 = hlfir.declare %[[VAL_0]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[VAL_2:.*]] = fir.address_of(@_QFEr) : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: %[[VAL_3:.*]]:2 = hlfir.declare %[[VAL_2]] {fortran_attrs = {{.*}}<pointer>, uniq_name = "_QFEr"} : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>) -> (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>)
+! CHECK: %[[VAL_4:.*]] = arith.constant false
+! CHECK: %[[VAL_5:.*]] = fir.absent !fir.box<none>
+! CHECK: %[[VAL_6:.*]] = fir.address_of(@_QQclX2c66ec21897efc91789c8c209a6eb0c0) : !fir.ref<!fir.char<1,89>>
+! CHECK: %[[VAL_7:.*]] = arith.constant 8 : i32
+! CHECK: %[[VAL_8:.*]] = fir.zero_bits !fir.ptr<!fir.array<?xi32>>
+! CHECK: %[[VAL_9:.*]] = arith.constant 0 : index
+! CHECK: %[[VAL_10:.*]] = fir.shape %[[VAL_9]] : (index) -> !fir.shape<1>
+! CHECK: %[[VAL_11:.*]] = fir.embox %[[VAL_8]](%[[VAL_10]]) : (!fir.ptr<!fir.array<?xi32>>, !fir.shape<1>) -> !fir.box<!fir.ptr<!fir.array<?xi32>>>
+! CHECK: fir.store %[[VAL_11]] to %[[VAL_3]]#1 : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: %[[VAL_12:.*]] = arith.constant 1 : index
+! CHECK: %[[VAL_13:.*]] = arith.constant 2 : i32
+! CHECK: %[[VAL_14:.*]] = arith.constant 0 : i32
+! CHECK: %[[VAL_15:.*]] = fir.convert %[[VAL_3]]#1 : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>) -> !fir.ref<!fir.box<none>>
+! CHECK: %[[VAL_16:.*]] = fir.convert %[[VAL_12]] : (index) -> i64
+! CHECK: %[[VAL_17:.*]] = fir.convert %[[VAL_13]] : (i32) -> i64
+! CHECK: %[[VAL_18:.*]] = fir.call @_FortranAPointerSetBounds(%[[VAL_15]], %[[VAL_14]], %[[VAL_16]], %[[VAL_17]]) fastmath<contract> : (!fir.ref<!fir.box<none>>, i32, i64, i64) -> none
+! CHECK: %[[VAL_19:.*]] = fir.convert %[[VAL_3]]#1 : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>) -> !fir.ref<!fir.box<none>>
+! CHECK: %[[VAL_20:.*]] = fir.convert %[[VAL_6]] : (!fir.ref<!fir.char<1,89>>) -> !fir.ref<i8>
+! CHECK: %[[VAL_21:.*]] = fir.call @_FortranAPointerAllocate(%[[VAL_19]], %[[VAL_4]], %[[VAL_5]], %[[VAL_20]], %[[VAL_7]]) fastmath<contract> : (!fir.ref<!fir.box<none>>, i1, !fir.box<none>, !fir.ref<i8>, i32) -> i32
+! CHECK: omp.parallel {
+! CHECK: %[[VAL_22:.*]] = fir.alloca i32 {bindc_name = "i", pinned, uniq_name = "_QFEi"}
+! CHECK: %[[VAL_23:.*]]:2 = hlfir.declare %[[VAL_22]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[VAL_24:.*]] = arith.constant 0 : i32
+! CHECK: %[[VAL_25:.*]] = arith.constant 10 : i32
+! CHECK: %[[VAL_26:.*]] = arith.constant 1 : i32
+! CHECK: omp.wsloop reduction(byref @add_reduction_byref_box_ptr_Uxi32 %[[VAL_3]]#0 -> %[[VAL_27:.*]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>) {
+! CHECK: omp.loop_nest (%[[VAL_28:.*]]) : i32 = (%[[VAL_24]]) to (%[[VAL_25]]) inclusive step (%[[VAL_26]]) {
+! CHECK: %[[VAL_29:.*]]:2 = hlfir.declare %[[VAL_27]] {fortran_attrs = {{.*}}<pointer>, uniq_name = "_QFEr"} : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>) -> (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>)
+! CHECK: fir.store %[[VAL_28]] to %[[VAL_23]]#1 : !fir.ref<i32>
+! CHECK: %[[VAL_30:.*]] = fir.load %[[VAL_23]]#0 : !fir.ref<i32>
+! CHECK: %[[VAL_31:.*]] = fir.load %[[VAL_29]]#0 : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: %[[VAL_32:.*]] = arith.constant 1 : index
+! CHECK: %[[VAL_33:.*]] = hlfir.designate %[[VAL_31]] (%[[VAL_32]]) : (!fir.box<!fir.ptr<!fir.array<?xi32>>>, index) -> !fir.ref<i32>
+! CHECK: hlfir.assign %[[VAL_30]] to %[[VAL_33]] : i32, !fir.ref<i32>
+! CHECK: %[[VAL_34:.*]] = fir.load %[[VAL_23]]#0 : !fir.ref<i32>
+! CHECK: %[[VAL_35:.*]] = arith.constant 0 : i32
+! CHECK: %[[VAL_36:.*]] = arith.subi %[[VAL_35]], %[[VAL_34]] : i32
+! CHECK: %[[VAL_37:.*]] = fir.load %[[VAL_29]]#0 : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
+! CHECK: %[[VAL_38:.*]] = arith.constant 2 : index
+! CHECK: %[[VAL_39:.*]] = hlfir.designate %[[VAL_37]] (%[[VAL_38]]) : (!fir.box<!fir.ptr<!fir.array<?xi32>>>, index) -> !fir.ref<i32>
+! CHECK: hlfir.assign %[[VAL_36]] to %[[VAL_39]] : i32, !fir.ref<i32>
+! CHECK: omp.yield
+! CHECK: }
+! CHECK: omp.terminator
+! CHECK: }
+! CHECK: omp.terminator
+! CHECK: }
diff --git a/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90 b/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
new file mode 100644
index 0000000000000..910d2ed0188d8
--- /dev/null
+++ b/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
@@ -0,0 +1,110 @@
+! RUN: bbc -emit-hlfir -fopenmp -o - %s | FileCheck %s
+! RUN: %flang_fc1 -emit-hlfir -fopenmp -o - %s | FileCheck %s
+
+program reduce_pointer
+ integer, pointer :: v
+ integer i
+
+ allocate(v)
+ v = 0
+
+ !$omp parallel do private(i) reduction(+:v)
+ do i = 1, 5
+ v = v + 42
+ end do
+ !$omp end parallel do
+
+ print *,v
+ deallocate(v)
+end program
+
+! CHECK-LABEL: omp.declare_reduction @add_reduction_byref_box_ptr_i32 : !fir.ref<!fir.box<!fir.ptr<i32>>> init {
+! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.ptr<i32>>>):
+! CHECK: %[[VAL_1:.*]] = arith.constant 0 : i32
+! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_3:.*]] = fir.alloca !fir.box<!fir.ptr<i32>>
+! CHECK: %[[VAL_4:.*]] = fir.box_addr %[[VAL_2]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
+! CHECK: %[[VAL_5:.*]] = fir.convert %[[VAL_4]] : (!fir.ptr<i32>) -> i64
+! CHECK: %[[VAL_6:.*]] = arith.constant 0 : i64
+! CHECK: %[[VAL_7:.*]] = arith.cmpi eq, %[[VAL_5]], %[[VAL_6]] : i64
+! CHECK: fir.if %[[VAL_7]] {
+! CHECK: %[[VAL_8:.*]] = fir.embox %[[VAL_4]] : (!fir.ptr<i32>) -> !fir.box<!fir.ptr<i32>>
+! CHECK: fir.store %[[VAL_8]] to %[[VAL_3]] : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: } else {
+! CHECK: %[[VAL_9:.*]] = fir.allocmem i32
+! CHECK: fir.store %[[VAL_1]] to %[[VAL_9]] : !fir.heap<i32>
+! CHECK: %[[VAL_10:.*]] = fir.embox %[[VAL_9]] : (!fir.heap<i32>) -> !fir.box<!fir.ptr<i32>>
+! CHECK: fir.store %[[VAL_10]] to %[[VAL_3]] : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: }
+! CHECK: omp.yield(%[[VAL_3]] : !fir.ref<!fir.box<!fir.ptr<i32>>>)
+! CHECK-LABEL: } combiner {
+! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.ptr<i32>>>, %[[VAL_1:.*]]: !fir.ref<!fir.box<!fir.ptr<i32>>>):
+! CHECK: %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_3:.*]] = fir.load %[[VAL_1]] : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_4:.*]] = fir.box_addr %[[VAL_2]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
+! CHECK: %[[VAL_5:.*]] = fir.box_addr %[[VAL_3]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
+! CHECK: %[[VAL_6:.*]] = fir.load %[[VAL_4]] : !fir.ptr<i32>
+! CHECK: %[[VAL_7:.*]] = fir.load %[[VAL_5]] : !fir.ptr<i32>
+! CHECK: %[[VAL_8:.*]] = arith.addi %[[VAL_6]], %[[VAL_7]] : i32
+! CHECK: fir.store %[[VAL_8]] to %[[VAL_4]] : !fir.ptr<i32>
+! CHECK: omp.yield(%[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<i32>>>)
+! CHECK-LABEL: } cleanup {
+! CHECK: ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.ptr<i32>>>):
+! CHECK: %[[VAL_1:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_2:.*]] = fir.box_addr %[[VAL_1]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
+! CHECK: %[[VAL_3:.*]] = fir.convert %[[VAL_2]] : (!fir.ptr<i32>) -> i64
+! CHECK: %[[VAL_4:.*]] = arith.constant 0 : i64
+! CHECK: %[[VAL_5:.*]] = arith.cmpi ne, %[[VAL_3]], %[[VAL_4]] : i64
+! CHECK: fir.if %[[VAL_5]] {
+! CHECK: %[[VAL_6:.*]] = fir.convert %[[VAL_2]] : (!fir.ptr<i32>) -> !fir.heap<i32>
+! CHECK: fir.freemem %[[VAL_6]] : !fir.heap<i32>
+! CHECK: }
+! CHECK: omp.yield
+! CHECK: }
+
+! CHECK-LABEL: func.func @_QQmain() attributes {fir.bindc_name = "reduce_pointer"} {
+! CHECK: %[[VAL_0:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFEi"}
+! CHECK: %[[VAL_1:.*]]:2 = hlfir.declare %[[VAL_0]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[VAL_2:.*]] = fir.alloca !fir.box<!fir.ptr<i32>> {bindc_name = "v", uniq_name = "_QFEv"}
+! CHECK: %[[VAL_3:.*]] = fir.zero_bits !fir.ptr<i32>
+! CHECK: %[[VAL_4:.*]] = fir.embox %[[VAL_3]] : (!fir.ptr<i32>) -> !fir.box<!fir.ptr<i32>>
+! CHECK: fir.store %[[VAL_4]] to %[[VAL_2]] : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_5:.*]]:2 = hlfir.declare %[[VAL_2]] {fortran_attrs = {{.*}}<pointer>, uniq_name = "_QFEv"} : (!fir.ref<!fir.box<!fir.ptr<i32>>>) -> (!fir.ref<!fir.box<!fir.ptr<i32>>>, !fir.ref<!fir.box<!fir.ptr<i32>>>)
+! CHECK: %[[VAL_6:.*]] = arith.constant false
+! CHECK: %[[VAL_7:.*]] = fir.absent !fir.box<none>
+! CHECK: %[[VAL_8:.*]] = fir.address_of(@_QQclXdaff7010de45a2c5c67f814ad3c8b2de) : !fir.ref<!fir.char<1,81>>
+! CHECK: %[[VAL_9:.*]] = arith.constant 8 : i32
+! CHECK: %[[VAL_10:.*]] = fir.zero_bits !fir.ptr<i32>
+! CHECK: %[[VAL_11:.*]] = fir.embox %[[VAL_10]] : (!fir.ptr<i32>) -> !fir.box<!fir.ptr<i32>>
+! CHECK: fir.store %[[VAL_11]] to %[[VAL_5]]#1 : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_12:.*]] = fir.convert %[[VAL_5]]#1 : (!fir.ref<!fir.box<!fir.ptr<i32>>>) -> !fir.ref<!fir.box<none>>
+! CHECK: %[[VAL_13:.*]] = fir.convert %[[VAL_8]] : (!fir.ref<!fir.char<1,81>>) -> !fir.ref<i8>
+! CHECK: %[[VAL_14:.*]] = fir.call @_FortranAPointerAllocate(%[[VAL_12]], %[[VAL_6]], %[[VAL_7]], %[[VAL_13]], %[[VAL_9]]) fastmath<contract> : (!fir.ref<!fir.box<none>>, i1, !fir.box<none>, !fir.ref<i8>, i32) -> i32
+! CHECK: %[[VAL_15:.*]] = arith.constant 0 : i32
+! CHECK: %[[VAL_16:.*]] = fir.load %[[VAL_5]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_17:.*]] = fir.box_addr %[[VAL_16]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
+! CHECK: hlfir.assign %[[VAL_15]] to %[[VAL_17]] : i32, !fir.ptr<i32>
+! CHECK: omp.parallel {
+! CHECK: %[[VAL_18:.*]] = fir.alloca i32 {bindc_name = "i", pinned, uniq_name = "_QFEi"}
+! CHECK: %[[VAL_19:.*]]:2 = hlfir.declare %[[VAL_18]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[VAL_20:.*]] = arith.constant 1 : i32
+! CHECK: %[[VAL_21:.*]] = arith.constant 5 : i32
+! CHECK: %[[VAL_22:.*]] = arith.constant 1 : i32
+! CHECK: omp.wsloop reduction(byref @add_reduction_byref_box_ptr_i32 %[[VAL_5]]#0 -> %[[VAL_23:.*]] : !fir.ref<!fir.box<!fir.ptr<i32>>>) {
+! CHECK: omp.loop_nest (%[[VAL_24:.*]]) : i32 = (%[[VAL_20]]) to (%[[VAL_21]]) inclusive step (%[[VAL_22]]) {
+! CHECK: %[[VAL_25:.*]]:2 = hlfir.declare %[[VAL_23]] {fortran_attrs = {{.*}}<pointer>, uniq_name = "_QFEv"} : (!fir.ref<!fir.box<!fir.ptr<i32>>>) -> (!fir.ref<!fir.box<!fir.ptr<i32>>>, !fir.ref<!fir.box<!fir.ptr<i32>>>)
+! CHECK: fir.store %[[VAL_24]] to %[[VAL_19]]#1 : !fir.ref<i32>
+! CHECK: %[[VAL_26:.*]] = fir.load %[[VAL_25]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_27:.*]] = fir.box_addr %[[VAL_26]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
+! CHECK: %[[VAL_28:.*]] = fir.load %[[VAL_27]] : !fir.ptr<i32>
+! CHECK: %[[VAL_29:.*]] = arith.constant 42 : i32
+! CHECK: %[[VAL_30:.*]] = arith.addi %[[VAL_28]], %[[VAL_29]] : i32
+! CHECK: %[[VAL_31:.*]] = fir.load %[[VAL_25]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
+! CHECK: %[[VAL_32:.*]] = fir.box_addr %[[VAL_31]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
+! CHECK: hlfir.assign %[[VAL_30]] to %[[VAL_32]] : i32, !fir.ptr<i32>
+! CHECK: omp.yield
+! CHECK: }
+! CHECK: omp.terminator
+! CHECK: }
+! CHECK: omp.terminator
+! CHECK: }
>From 0b8abe81628288b4bdad03d62cb0d47a825ebba1 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles at arm.com>
Date: Wed, 12 Jun 2024 09:37:16 +0000
Subject: [PATCH 2/5] (Hopefully) fix test failure
I couldn't reproduce it locally
---
flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90 | 2 +-
flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90 | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90 b/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
index dfdc50d0c116d..571647ffce5f2 100644
--- a/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
+++ b/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
@@ -84,7 +84,7 @@ program reduce
! CHECK: %[[VAL_3:.*]]:2 = hlfir.declare %[[VAL_2]] {fortran_attrs = {{.*}}<pointer>, uniq_name = "_QFEr"} : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>) -> (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>)
! CHECK: %[[VAL_4:.*]] = arith.constant false
! CHECK: %[[VAL_5:.*]] = fir.absent !fir.box<none>
-! CHECK: %[[VAL_6:.*]] = fir.address_of(@_QQclX2c66ec21897efc91789c8c209a6eb0c0) : !fir.ref<!fir.char<1,89>>
+! CHECK: %[[VAL_6:.*]] = fir.address_of({{.*}}) : !fir.ref<!fir.char<1,89>>
! CHECK: %[[VAL_7:.*]] = arith.constant 8 : i32
! CHECK: %[[VAL_8:.*]] = fir.zero_bits !fir.ptr<!fir.array<?xi32>>
! CHECK: %[[VAL_9:.*]] = arith.constant 0 : index
diff --git a/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90 b/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
index 910d2ed0188d8..c810875ab3e2a 100644
--- a/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
+++ b/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
@@ -72,7 +72,7 @@ program reduce_pointer
! CHECK: %[[VAL_5:.*]]:2 = hlfir.declare %[[VAL_2]] {fortran_attrs = {{.*}}<pointer>, uniq_name = "_QFEv"} : (!fir.ref<!fir.box<!fir.ptr<i32>>>) -> (!fir.ref<!fir.box<!fir.ptr<i32>>>, !fir.ref<!fir.box<!fir.ptr<i32>>>)
! CHECK: %[[VAL_6:.*]] = arith.constant false
! CHECK: %[[VAL_7:.*]] = fir.absent !fir.box<none>
-! CHECK: %[[VAL_8:.*]] = fir.address_of(@_QQclXdaff7010de45a2c5c67f814ad3c8b2de) : !fir.ref<!fir.char<1,81>>
+! CHECK: %[[VAL_8:.*]] = fir.address_of({{.*}}) : !fir.ref<!fir.char<1,81>>
! CHECK: %[[VAL_9:.*]] = arith.constant 8 : i32
! CHECK: %[[VAL_10:.*]] = fir.zero_bits !fir.ptr<i32>
! CHECK: %[[VAL_11:.*]] = fir.embox %[[VAL_10]] : (!fir.ptr<i32>) -> !fir.box<!fir.ptr<i32>>
>From 238a1aa967349a64b2e87643e75544210e544f88 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles at arm.com>
Date: Thu, 13 Jun 2024 12:31:40 +0000
Subject: [PATCH 3/5] Fix assert message
---
flang/lib/Lower/OpenMP/ReductionProcessor.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
index b32036f4b5f4a..237f9cdf22a1b 100644
--- a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
@@ -593,7 +593,7 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
createReductionCleanupRegion(builder, loc, reductionDecl);
} else {
assert(!isAllocatableOrPointer &&
- "Allocatable arrays must be heap allocated");
+ "Pointer-like arrays must be heap allocated");
}
// Put the temporary inside of a box:
>From 26e8d73dd4b7fec705181468d567bd704b844076 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles at arm.com>
Date: Thu, 13 Jun 2024 12:33:20 +0000
Subject: [PATCH 4/5] Second attempt to fix test failure
---
flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90 | 2 +-
flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90 | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90 b/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
index 571647ffce5f2..643a89b40c695 100644
--- a/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
+++ b/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
@@ -84,7 +84,7 @@ program reduce
! CHECK: %[[VAL_3:.*]]:2 = hlfir.declare %[[VAL_2]] {fortran_attrs = {{.*}}<pointer>, uniq_name = "_QFEr"} : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>) -> (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>)
! CHECK: %[[VAL_4:.*]] = arith.constant false
! CHECK: %[[VAL_5:.*]] = fir.absent !fir.box<none>
-! CHECK: %[[VAL_6:.*]] = fir.address_of({{.*}}) : !fir.ref<!fir.char<1,89>>
+! CHECK: %[[VAL_6:.*]] = fir.address_of(
! CHECK: %[[VAL_7:.*]] = arith.constant 8 : i32
! CHECK: %[[VAL_8:.*]] = fir.zero_bits !fir.ptr<!fir.array<?xi32>>
! CHECK: %[[VAL_9:.*]] = arith.constant 0 : index
diff --git a/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90 b/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
index c810875ab3e2a..3fd319255d2af 100644
--- a/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
+++ b/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
@@ -72,7 +72,7 @@ program reduce_pointer
! CHECK: %[[VAL_5:.*]]:2 = hlfir.declare %[[VAL_2]] {fortran_attrs = {{.*}}<pointer>, uniq_name = "_QFEv"} : (!fir.ref<!fir.box<!fir.ptr<i32>>>) -> (!fir.ref<!fir.box<!fir.ptr<i32>>>, !fir.ref<!fir.box<!fir.ptr<i32>>>)
! CHECK: %[[VAL_6:.*]] = arith.constant false
! CHECK: %[[VAL_7:.*]] = fir.absent !fir.box<none>
-! CHECK: %[[VAL_8:.*]] = fir.address_of({{.*}}) : !fir.ref<!fir.char<1,81>>
+! CHECK: %[[VAL_8:.*]] = fir.address_of(
! CHECK: %[[VAL_9:.*]] = arith.constant 8 : i32
! CHECK: %[[VAL_10:.*]] = fir.zero_bits !fir.ptr<i32>
! CHECK: %[[VAL_11:.*]] = fir.embox %[[VAL_10]] : (!fir.ptr<i32>) -> !fir.box<!fir.ptr<i32>>
>From a75943e86e3956422ebc2900dbb8fec0db283f1f Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles at arm.com>
Date: Thu, 13 Jun 2024 13:17:16 +0000
Subject: [PATCH 5/5] More attempted test fixes
---
flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90 | 2 +-
flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90 | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90 b/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
index 643a89b40c695..2c2f60cb72c9a 100644
--- a/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
+++ b/flang/test/Lower/OpenMP/parallel-reduction-pointer-array.f90
@@ -99,7 +99,7 @@ program reduce
! CHECK: %[[VAL_17:.*]] = fir.convert %[[VAL_13]] : (i32) -> i64
! CHECK: %[[VAL_18:.*]] = fir.call @_FortranAPointerSetBounds(%[[VAL_15]], %[[VAL_14]], %[[VAL_16]], %[[VAL_17]]) fastmath<contract> : (!fir.ref<!fir.box<none>>, i32, i64, i64) -> none
! CHECK: %[[VAL_19:.*]] = fir.convert %[[VAL_3]]#1 : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>) -> !fir.ref<!fir.box<none>>
-! CHECK: %[[VAL_20:.*]] = fir.convert %[[VAL_6]] : (!fir.ref<!fir.char<1,89>>) -> !fir.ref<i8>
+! CHECK: %[[VAL_20:.*]] = fir.convert %[[VAL_6]] : (!fir.ref<!fir.char<{{.*}}>>) -> !fir.ref<i8>
! CHECK: %[[VAL_21:.*]] = fir.call @_FortranAPointerAllocate(%[[VAL_19]], %[[VAL_4]], %[[VAL_5]], %[[VAL_20]], %[[VAL_7]]) fastmath<contract> : (!fir.ref<!fir.box<none>>, i1, !fir.box<none>, !fir.ref<i8>, i32) -> i32
! CHECK: omp.parallel {
! CHECK: %[[VAL_22:.*]] = fir.alloca i32 {bindc_name = "i", pinned, uniq_name = "_QFEi"}
diff --git a/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90 b/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
index 3fd319255d2af..aab6efbcbc5fe 100644
--- a/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
+++ b/flang/test/Lower/OpenMP/wsloop-reduction-pointer.f90
@@ -78,7 +78,7 @@ program reduce_pointer
! CHECK: %[[VAL_11:.*]] = fir.embox %[[VAL_10]] : (!fir.ptr<i32>) -> !fir.box<!fir.ptr<i32>>
! CHECK: fir.store %[[VAL_11]] to %[[VAL_5]]#1 : !fir.ref<!fir.box<!fir.ptr<i32>>>
! CHECK: %[[VAL_12:.*]] = fir.convert %[[VAL_5]]#1 : (!fir.ref<!fir.box<!fir.ptr<i32>>>) -> !fir.ref<!fir.box<none>>
-! CHECK: %[[VAL_13:.*]] = fir.convert %[[VAL_8]] : (!fir.ref<!fir.char<1,81>>) -> !fir.ref<i8>
+! CHECK: %[[VAL_13:.*]] = fir.convert %[[VAL_8]] : (!fir.ref<!fir.char<{{.*}}>>) -> !fir.ref<i8>
! CHECK: %[[VAL_14:.*]] = fir.call @_FortranAPointerAllocate(%[[VAL_12]], %[[VAL_6]], %[[VAL_7]], %[[VAL_13]], %[[VAL_9]]) fastmath<contract> : (!fir.ref<!fir.box<none>>, i1, !fir.box<none>, !fir.ref<i8>, i32) -> i32
! CHECK: %[[VAL_15:.*]] = arith.constant 0 : i32
! CHECK: %[[VAL_16:.*]] = fir.load %[[VAL_5]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
More information about the flang-commits
mailing list