[flang] Propagate volatile on openmp reduction vars #142182
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
In the openmp reduction clause lowering, reduction variables"'" reference types were not propagating the volatility of the original variable's type. Add a test and address cases of this in ReductionProcessor.
|
@llvm/pr-subscribers-flang-openmp Author: Asher Mancinelli (ashermancinelli) ChangesIn the openmp reduction clause lowering, reduction variables' reference types were not propagating the volatility of the original variable's type. Add a test and address cases of this in ReductionProcessor. I'm also looking for other cases of this in the openmp bridge, but haven't found any yet. Full diff: https://github.com/llvm/llvm-project/pull/142182.diff 2 Files Affected:
diff --git a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
index b8aa0deb42dd6..7ef0f2a0ef7c5 100644
--- a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
@@ -389,7 +389,8 @@ static void genBoxCombiner(fir::FirOpBuilder &builder, mlir::Location loc,
hlfir::LoopNest nest = hlfir::genLoopNest(
loc, builder, shapeShift.getExtents(), /*isUnordered=*/true);
builder.setInsertionPointToStart(nest.body);
- mlir::Type refTy = fir::ReferenceType::get(seqTy.getEleTy());
+ const bool seqIsVolatile = fir::isa_volatile_type(seqTy.getEleTy());
+ mlir::Type refTy = fir::ReferenceType::get(seqTy.getEleTy(), seqIsVolatile);
auto lhsEleAddr = builder.create<fir::ArrayCoorOp>(
loc, refTy, lhs, shapeShift, /*slice=*/mlir::Value{},
nest.oneBasedIndices, /*typeparms=*/mlir::ValueRange{});
@@ -659,7 +660,8 @@ void ReductionProcessor::processReductionArguments(
assert(fir::isa_ref_type(symVal.getType()) &&
"reduction input var is passed by reference");
mlir::Type elementType = fir::dyn_cast_ptrEleTy(symVal.getType());
- mlir::Type refTy = fir::ReferenceType::get(elementType);
+ const bool symIsVolatile = fir::isa_volatile_type(symVal.getType());
+ mlir::Type refTy = fir::ReferenceType::get(elementType, symIsVolatile);
reductionVars.push_back(
builder.createConvert(currentLocation, refTy, symVal));
diff --git a/flang/test/Lower/volatile-openmp1.f90 b/flang/test/Lower/volatile-openmp1.f90
new file mode 100644
index 0000000000000..163db953b6b80
--- /dev/null
+++ b/flang/test/Lower/volatile-openmp1.f90
@@ -0,0 +1,46 @@
+! RUN: bbc --strict-fir-volatile-verifier -fopenmp %s -o - | FileCheck %s
+program main
+implicit none
+integer,volatile::a
+integer::n,i
+a=0
+n=1000
+!$omp parallel
+!$omp do reduction(+:a)
+ do i=1,n
+ a=a+1
+ end do
+!$omp end parallel
+end program
+
+! CHECK-LABEL: func.func @_QQmain() attributes {fir.bindc_name = "main"} {
+! CHECK: %[[VAL_0:.*]] = arith.constant 1 : i32
+! CHECK: %[[VAL_1:.*]] = arith.constant 1000 : i32
+! CHECK: %[[VAL_2:.*]] = arith.constant 0 : i32
+! CHECK: %[[VAL_3:.*]] = fir.dummy_scope : !fir.dscope
+! CHECK: %[[VAL_4:.*]] = fir.alloca i32 {bindc_name = "a", uniq_name = "_QFEa"}
+! CHECK: %[[VAL_5:.*]] = fir.volatile_cast %[[VAL_4]] : (!fir.ref<i32>) -> !fir.ref<i32, volatile>
+! CHECK: %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_5]] {fortran_attrs = #fir.var_attrs<volatile>, uniq_name = "_QFEa"} : (!fir.ref<i32, volatile>) -> (!fir.ref<i32, volatile>, !fir.ref<i32, volatile>)
+! CHECK: %[[VAL_7:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFEi"}
+! CHECK: %[[VAL_8:.*]]:2 = hlfir.declare %[[VAL_7]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[VAL_9:.*]] = fir.alloca i32 {bindc_name = "n", uniq_name = "_QFEn"}
+! CHECK: %[[VAL_10:.*]]:2 = hlfir.declare %[[VAL_9]] {uniq_name = "_QFEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: hlfir.assign %[[VAL_2]] to %[[VAL_6]]#0 : i32, !fir.ref<i32, volatile>
+! CHECK: hlfir.assign %[[VAL_1]] to %[[VAL_10]]#0 : i32, !fir.ref<i32>
+! CHECK: omp.parallel {
+! CHECK: %[[VAL_11:.*]] = fir.load %[[VAL_10]]#0 : !fir.ref<i32>
+! CHECK: omp.wsloop private(@_QFEi_private_i32 %[[VAL_8]]#0 -> %[[VAL_12:.*]] : !fir.ref<i32>) reduction(@add_reduction_i32 %[[VAL_6]]#0 -> %[[VAL_13:.*]] : !fir.ref<i32, volatile>) {
+! CHECK: omp.loop_nest (%[[VAL_14:.*]]) : i32 = (%[[VAL_0]]) to (%[[VAL_11]]) inclusive step (%[[VAL_0]]) {
+! CHECK: %[[VAL_15:.*]]:2 = hlfir.declare %[[VAL_12]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[VAL_16:.*]]:2 = hlfir.declare %[[VAL_13]] {fortran_attrs = #fir.var_attrs<volatile>, uniq_name = "_QFEa"} : (!fir.ref<i32, volatile>) -> (!fir.ref<i32, volatile>, !fir.ref<i32, volatile>)
+! CHECK: hlfir.assign %[[VAL_14]] to %[[VAL_15]]#0 : i32, !fir.ref<i32>
+! CHECK: %[[VAL_17:.*]] = fir.load %[[VAL_16]]#0 : !fir.ref<i32, volatile>
+! CHECK: %[[VAL_18:.*]] = arith.addi %[[VAL_17]], %[[VAL_0]] : i32
+! CHECK: hlfir.assign %[[VAL_18]] to %[[VAL_16]]#0 : i32, !fir.ref<i32, volatile>
+! CHECK: omp.yield
+! CHECK: }
+! CHECK: }
+! CHECK: omp.terminator
+! CHECK: }
+! CHECK: return
+! CHECK: }
|
|
@llvm/pr-subscribers-flang-fir-hlfir Author: Asher Mancinelli (ashermancinelli) ChangesIn the openmp reduction clause lowering, reduction variables' reference types were not propagating the volatility of the original variable's type. Add a test and address cases of this in ReductionProcessor. I'm also looking for other cases of this in the openmp bridge, but haven't found any yet. Full diff: https://github.com/llvm/llvm-project/pull/142182.diff 2 Files Affected:
diff --git a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
index b8aa0deb42dd6..7ef0f2a0ef7c5 100644
--- a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
@@ -389,7 +389,8 @@ static void genBoxCombiner(fir::FirOpBuilder &builder, mlir::Location loc,
hlfir::LoopNest nest = hlfir::genLoopNest(
loc, builder, shapeShift.getExtents(), /*isUnordered=*/true);
builder.setInsertionPointToStart(nest.body);
- mlir::Type refTy = fir::ReferenceType::get(seqTy.getEleTy());
+ const bool seqIsVolatile = fir::isa_volatile_type(seqTy.getEleTy());
+ mlir::Type refTy = fir::ReferenceType::get(seqTy.getEleTy(), seqIsVolatile);
auto lhsEleAddr = builder.create<fir::ArrayCoorOp>(
loc, refTy, lhs, shapeShift, /*slice=*/mlir::Value{},
nest.oneBasedIndices, /*typeparms=*/mlir::ValueRange{});
@@ -659,7 +660,8 @@ void ReductionProcessor::processReductionArguments(
assert(fir::isa_ref_type(symVal.getType()) &&
"reduction input var is passed by reference");
mlir::Type elementType = fir::dyn_cast_ptrEleTy(symVal.getType());
- mlir::Type refTy = fir::ReferenceType::get(elementType);
+ const bool symIsVolatile = fir::isa_volatile_type(symVal.getType());
+ mlir::Type refTy = fir::ReferenceType::get(elementType, symIsVolatile);
reductionVars.push_back(
builder.createConvert(currentLocation, refTy, symVal));
diff --git a/flang/test/Lower/volatile-openmp1.f90 b/flang/test/Lower/volatile-openmp1.f90
new file mode 100644
index 0000000000000..163db953b6b80
--- /dev/null
+++ b/flang/test/Lower/volatile-openmp1.f90
@@ -0,0 +1,46 @@
+! RUN: bbc --strict-fir-volatile-verifier -fopenmp %s -o - | FileCheck %s
+program main
+implicit none
+integer,volatile::a
+integer::n,i
+a=0
+n=1000
+!$omp parallel
+!$omp do reduction(+:a)
+ do i=1,n
+ a=a+1
+ end do
+!$omp end parallel
+end program
+
+! CHECK-LABEL: func.func @_QQmain() attributes {fir.bindc_name = "main"} {
+! CHECK: %[[VAL_0:.*]] = arith.constant 1 : i32
+! CHECK: %[[VAL_1:.*]] = arith.constant 1000 : i32
+! CHECK: %[[VAL_2:.*]] = arith.constant 0 : i32
+! CHECK: %[[VAL_3:.*]] = fir.dummy_scope : !fir.dscope
+! CHECK: %[[VAL_4:.*]] = fir.alloca i32 {bindc_name = "a", uniq_name = "_QFEa"}
+! CHECK: %[[VAL_5:.*]] = fir.volatile_cast %[[VAL_4]] : (!fir.ref<i32>) -> !fir.ref<i32, volatile>
+! CHECK: %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_5]] {fortran_attrs = #fir.var_attrs<volatile>, uniq_name = "_QFEa"} : (!fir.ref<i32, volatile>) -> (!fir.ref<i32, volatile>, !fir.ref<i32, volatile>)
+! CHECK: %[[VAL_7:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFEi"}
+! CHECK: %[[VAL_8:.*]]:2 = hlfir.declare %[[VAL_7]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[VAL_9:.*]] = fir.alloca i32 {bindc_name = "n", uniq_name = "_QFEn"}
+! CHECK: %[[VAL_10:.*]]:2 = hlfir.declare %[[VAL_9]] {uniq_name = "_QFEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: hlfir.assign %[[VAL_2]] to %[[VAL_6]]#0 : i32, !fir.ref<i32, volatile>
+! CHECK: hlfir.assign %[[VAL_1]] to %[[VAL_10]]#0 : i32, !fir.ref<i32>
+! CHECK: omp.parallel {
+! CHECK: %[[VAL_11:.*]] = fir.load %[[VAL_10]]#0 : !fir.ref<i32>
+! CHECK: omp.wsloop private(@_QFEi_private_i32 %[[VAL_8]]#0 -> %[[VAL_12:.*]] : !fir.ref<i32>) reduction(@add_reduction_i32 %[[VAL_6]]#0 -> %[[VAL_13:.*]] : !fir.ref<i32, volatile>) {
+! CHECK: omp.loop_nest (%[[VAL_14:.*]]) : i32 = (%[[VAL_0]]) to (%[[VAL_11]]) inclusive step (%[[VAL_0]]) {
+! CHECK: %[[VAL_15:.*]]:2 = hlfir.declare %[[VAL_12]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[VAL_16:.*]]:2 = hlfir.declare %[[VAL_13]] {fortran_attrs = #fir.var_attrs<volatile>, uniq_name = "_QFEa"} : (!fir.ref<i32, volatile>) -> (!fir.ref<i32, volatile>, !fir.ref<i32, volatile>)
+! CHECK: hlfir.assign %[[VAL_14]] to %[[VAL_15]]#0 : i32, !fir.ref<i32>
+! CHECK: %[[VAL_17:.*]] = fir.load %[[VAL_16]]#0 : !fir.ref<i32, volatile>
+! CHECK: %[[VAL_18:.*]] = arith.addi %[[VAL_17]], %[[VAL_0]] : i32
+! CHECK: hlfir.assign %[[VAL_18]] to %[[VAL_16]]#0 : i32, !fir.ref<i32, volatile>
+! CHECK: omp.yield
+! CHECK: }
+! CHECK: }
+! CHECK: omp.terminator
+! CHECK: }
+! CHECK: return
+! CHECK: }
|
jeanPerier
approved these changes
Jun 2, 2025
sallto
pushed a commit
to sallto/llvm-project
that referenced
this pull request
Jun 3, 2025
In the openmp reduction clause lowering, reduction variables' reference types were not propagating the volatility of the original variable's type. Add a test and address cases of this in ReductionProcessor.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In the openmp reduction clause lowering, reduction variables' reference types were not propagating the volatility of the original variable's type. Add a test and address cases of this in ReductionProcessor.
I'm also looking for other cases of this in the openmp bridge, but haven't found any yet.