[OpenACC] Fix issue with array-section type generation #162499
Merged
+6
−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
When an array section bound was in a list of bound-types we improperly
generated the list of types of the bounds because
getPointeeOrElementType gets the LOWEST level of type (that is, digs
through ALL array types to get to the base-est of types) when what
we really wanted was 1 layer of pointer/array removed.
This patch fixes it and adds a test that showed the problem by
re-ordering the existing ones. This wasn't previously obvious by
chance, since the 'array-index-only' variants ended up generating the
recipe, and not the bounds.
erichkeane
requested review from
lanza,
bcardosolopes,
xlauko and
andykaylor
as code owners
llvmbot
added
clang
|
@llvm/pr-subscribers-clangir @llvm/pr-subscribers-clang Author: Erich Keane (erichkeane) ChangesWhen an array section bound was in a list of bound-types we improperly generated the list of types of the bounds because This patch fixes it and adds a test that showed the problem by re-ordering the existing ones. This wasn't previously obvious by chance, since the 'array-index-only' variants ended up generating the recipe, and not the bounds. Full diff: https://github.com/llvm/llvm-project/pull/162499.diff 2 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenOpenACC.cpp b/clang/lib/CIR/CodeGen/CIRGenOpenACC.cpp
index a9af753381db3..4cf2237468afd 100644
--- a/clang/lib/CIR/CodeGen/CIRGenOpenACC.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenOpenACC.cpp
@@ -87,7 +87,10 @@ CIRGenFunction::getOpenACCDataOperandInfo(const Expr *e) {
if (const auto *section = dyn_cast<ArraySectionExpr>(curVarExpr)) {
QualType baseTy = ArraySectionExpr::getBaseOriginalType(
section->getBase()->IgnoreParenImpCasts());
- boundTypes.push_back(QualType(baseTy->getPointeeOrArrayElementType(), 0));
+ if (auto *at = getContext().getAsArrayType(baseTy))
+ boundTypes.push_back(at->getElementType());
+ else
+ boundTypes.push_back(baseTy->getPointeeType());
} else {
boundTypes.push_back(curVarExpr->getType());
}
diff --git a/clang/test/CIR/CodeGenOpenACC/private-clause-array-recipes-int.cpp b/clang/test/CIR/CodeGenOpenACC/private-clause-array-recipes-int.cpp
index e83e548cd138b..74cb567076f21 100644
--- a/clang/test/CIR/CodeGenOpenACC/private-clause-array-recipes-int.cpp
+++ b/clang/test/CIR/CodeGenOpenACC/private-clause-array-recipes-int.cpp
@@ -21,7 +21,7 @@ void do_things(unsigned A, unsigned B) {
;
T TwoArr[5][5];
-#pragma acc parallel private(TwoArr[B][B])
+#pragma acc parallel private(TwoArr[A:B][A:B])
// CHECK-NEXT: acc.private.recipe @privatization__Bcnt2__ZTSA5_A5_i : !cir.ptr<!cir.array<!cir.array<!s32i x 5> x 5>> init {
// CHECK-NEXT: ^bb0(%arg0: !cir.ptr<!cir.array<!cir.array<!s32i x 5> x 5>> {{.*}}, %[[BOUND1:.*]]: !acc.data_bounds_ty {{.*}}, %[[BOUND2:.*]]: !acc.data_bounds_ty {{.*}}):
// CHECK-NEXT: %[[TL_ALLOCA:.*]] = cir.alloca !cir.array<!cir.array<!s32i x 5> x 5>, !cir.ptr<!cir.array<!cir.array<!s32i x 5> x 5>>, ["openacc.private.init"] {alignment = 4 : i64}
@@ -30,7 +30,7 @@ void do_things(unsigned A, unsigned B) {
;
#pragma acc parallel private(TwoArr[B][A:B])
;
-#pragma acc parallel private(TwoArr[A:B][A:B])
+#pragma acc parallel private(TwoArr[B][B])
;
#pragma acc parallel private(TwoArr)
// CHECK-NEXT: acc.private.recipe @privatization__ZTSA5_A5_i : !cir.ptr<!cir.array<!cir.array<!s32i x 5> x 5>> init {
|
svkeerthy
pushed a commit
that referenced
this pull request
Oct 9, 2025
When an array section bound was in a list of bound-types we improperly
generated the list of types of the bounds because
getPointeeOrElementType gets the LOWEST level of type (that is, digs
through ALL array types to get to the base-est of types) when what
we really wanted was 1 layer of pointer/array removed.
This patch fixes it and adds a test that showed the problem by
re-ordering the existing ones. This wasn't previously obvious by chance,
since the 'array-index-only' variants ended up generating the recipe,
and not the bounds.
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.
When an array section bound was in a list of bound-types we improperly generated the list of types of the bounds because
getPointeeOrElementType gets the LOWEST level of type (that is, digs
through ALL array types to get to the base-est of types) when what
we really wanted was 1 layer of pointer/array removed.
This patch fixes it and adds a test that showed the problem by re-ordering the existing ones. This wasn't previously obvious by chance, since the 'array-index-only' variants ended up generating the recipe, and not the bounds.