[flang][OpenMP] Fix fir.convert in omp.atomic.update region #138397
Merged
+35
−3
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
|
@llvm/pr-subscribers-flang-fir-hlfir @llvm/pr-subscribers-flang-openmp Author: None (NimishMishra) ChangesRegion generation in omp.atomic.update currently emits a direct Fixes #138396 Full diff: https://github.com/llvm/llvm-project/pull/138397.diff 2 Files Affected:
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 47e7c266ff7d3..0910d0003356c 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -2833,9 +2833,23 @@ static void genAtomicUpdateStatement(
lower::StatementContext atomicStmtCtx;
mlir::Value rhsExpr = fir::getBase(converter.genExprValue(
*semantics::GetExpr(assignmentStmtExpr), atomicStmtCtx));
- mlir::Value convertResult =
- firOpBuilder.createConvert(currentLocation, varType, rhsExpr);
- firOpBuilder.create<mlir::omp::YieldOp>(currentLocation, convertResult);
+ mlir::Type exprType = fir::unwrapRefType(rhsExpr.getType());
+ if (fir::isa_complex(exprType) && !fir::isa_complex(varType)) {
+ // Emit an additional `ExtractValueOp` if the expression is of complex
+ // type
+ auto extract = firOpBuilder.create<fir::ExtractValueOp>(
+ currentLocation,
+ mlir::cast<mlir::ComplexType>(exprType).getElementType(), rhsExpr,
+ firOpBuilder.getArrayAttr(
+ firOpBuilder.getIntegerAttr(firOpBuilder.getIndexType(), 0)));
+ mlir::Value convertResult = firOpBuilder.create<fir::ConvertOp>(
+ currentLocation, varType, extract);
+ firOpBuilder.create<mlir::omp::YieldOp>(currentLocation, convertResult);
+ } else {
+ mlir::Value convertResult =
+ firOpBuilder.createConvert(currentLocation, varType, rhsExpr);
+ firOpBuilder.create<mlir::omp::YieldOp>(currentLocation, convertResult);
+ }
converter.resetExprOverrides();
}
firOpBuilder.setInsertionPointAfter(atomicUpdateOp);
diff --git a/flang/test/Lower/OpenMP/atomic-update.f90 b/flang/test/Lower/OpenMP/atomic-update.f90
index 31bf447006930..257ae8fb497ff 100644
--- a/flang/test/Lower/OpenMP/atomic-update.f90
+++ b/flang/test/Lower/OpenMP/atomic-update.f90
@@ -20,6 +20,8 @@ program OmpAtomicUpdate
!CHECK: %[[VAL_C_DECLARE:.*]]:2 = hlfir.declare %[[VAL_C_ADDRESS]] {{.*}}
!CHECK: %[[VAL_D_ADDRESS:.*]] = fir.address_of(@_QFEd) : !fir.ref<i32>
!CHECK: %[[VAL_D_DECLARE:.*]]:2 = hlfir.declare %[[VAL_D_ADDRESS]] {{.}}
+!CHECK: %[[VAL_G_ADDRESS:.*]] = fir.alloca complex<f32> {bindc_name = "g", uniq_name = "_QFEg"}
+!CHECK: %[[VAL_G_DECLARE:.*]]:2 = hlfir.declare %[[VAL_G_ADDRESS]] {uniq_name = "_QFEg"} : (!fir.ref<complex<f32>>) -> (!fir.ref<complex<f32>>, !fir.ref<complex<f32>>)
!CHECK: %[[VAL_i1_ALLOCA:.*]] = fir.alloca i8 {bindc_name = "i1", uniq_name = "_QFEi1"}
!CHECK: %[[VAL_i1_DECLARE:.*]]:2 = hlfir.declare %[[VAL_i1_ALLOCA]] {{.*}}
!CHECK: %[[VAL_c5:.*]] = arith.constant 5 : index
@@ -40,6 +42,7 @@ program OmpAtomicUpdate
integer, target :: c, d
integer(1) :: i1
integer, dimension(5) :: k
+ complex :: g
!CHECK: %[[EMBOX:.*]] = fir.embox %[[VAL_C_DECLARE]]#0 : (!fir.ref<i32>) -> !fir.box<!fir.ptr<i32>>
!CHECK: fir.store %[[EMBOX]] to %[[VAL_A_DECLARE]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
@@ -200,4 +203,19 @@ program OmpAtomicUpdate
!CHECK: }
!$omp atomic update
x = x + sum([ (y+2, y=1, z) ])
+
+!CHECK: %[[LOAD:.*]] = fir.load %[[VAL_G_DECLARE]]#0 : !fir.ref<complex<f32>>
+!CHECK: omp.atomic.update %[[VAL_W_DECLARE]]#0 : !fir.ref<i32> {
+!CHECK: ^bb0(%[[ARG:.*]]: i32):
+!CHECK: %[[CVT:.*]] = fir.convert %[[ARG]] : (i32) -> f32
+!CHECK: %[[CST:.*]] = arith.constant 0.000000e+00 : f32
+!CHECK: %[[UNDEF:.*]] = fir.undefined complex<f32>
+!CHECK: %[[IDX1:.*]] = fir.insert_value %[[UNDEF]], %[[CVT]], [0 : index] : (complex<f32>, f32) -> complex<f32>
+!CHECK: %[[IDX2:.*]] = fir.insert_value %[[IDX1]], %[[CST]], [1 : index] : (complex<f32>, f32) -> complex<f32>
+!CHECK: %[[ADD:.*]] = fir.addc %[[IDX2]], %[[LOAD]] {fastmath = #arith.fastmath<contract>} : complex<f32>
+!CHECK: %[[EXT:.*]] = fir.extract_value %[[ADD]], [0 : index] : (complex<f32>) -> f32
+!CHECK: %[[RESULT:.*]] = fir.convert %[[EXT]] : (f32) -> i32
+!CHECK: omp.yield(%[[RESULT]] : i32)
+ !$omp atomic update
+ w = w + g
end program OmpAtomicUpdate
|
Thirumalai-Shaktivel
approved these changes
May 6, 2025
GeorgeARM
pushed a commit
to GeorgeARM/llvm-project
that referenced
this pull request
May 7, 2025
) Region generation in omp.atomic.update currently emits a direct `fir.convert`. This crashes when the RHS expression involves complex type but the LHS variable is primitive type (say `f32`), since a `fir.convert` from `complex<f32>` to `f32` is emitted, which is illegal. This PR adds a conditional check to emit an additional `ExtractValueOp` in case RHS expression has a complex type. Fixes llvm#138396
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.
Region generation in omp.atomic.update currently emits a direct
fir.convert. This crashes when the RHS expression involves complex type but the LHS variable is primitive type (sayf32), since afir.convertfromcomplex<f32>tof32is emitted, which is illegal. This PR adds a conditional check to emit an additionalExtractValueOpin case RHS expression has a complex type.Fixes #138396