[clang] [Clang][OpenMP] Fix static,1 inner loop bound (PR #239868)
via cfe-commits
cfe-commits at lists.llvm.org
Fri Nov 6 17:13:49 PST 2026
llvmorg-github-actions[bot] wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-clang
Author: Amadeu Moya Sardà (amad00r)
<details>
<summary>Changes</summary>
The inner loop of a combined distribute parallel for compared its iteration variable directly with .previous.ub., whose parameter type is size_t. This could widen a signed i32 iteration variable to i64 and make the comparison unsigned.
Reuse the existing conversion of .previous.ub. to the loop upper-bound type when building the inner-loop condition. Update the affected OpenMP codegen checks and add a focused test for signed and unsigned i32 loops.
---
Patch is 61.65 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/239868.diff
11 Files Affected:
- (modified) clang/lib/Sema/SemaOpenMP.cpp (+1-1)
- (modified) clang/test/OpenMP/amdgcn_target_device_vla.cpp (+2-2)
- (modified) clang/test/OpenMP/amdgpu_target_with_aligned_attribute.c (+2-2)
- (added) clang/test/OpenMP/distribute_parallel_for_bound_type_codegen.c (+30)
- (modified) clang/test/OpenMP/nvptx_SPMD_codegen.cpp (+52-52)
- (modified) clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp (+3-3)
- (modified) clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp (+23-23)
- (modified) clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp (+9-9)
- (modified) clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp (+23-23)
- (modified) clang/test/OpenMP/target_teams_generic_loop_codegen.cpp (+2-2)
- (modified) clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp (+6-6)
``````````diff
diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index 6f4c71b118d590..baad51b6321f4f 100644
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -11417,7 +11417,7 @@ checkOpenMPLoop(OpenMPDirectiveKind DKind, Expr *CollapseLoopCountExpr,
// Build IV <= PrevUB or IV < PrevUB + 1 for unsigned IV to be used in
// parallel for is in combination with a distribute directive with
// schedule(static, 1)
- Expr *BoundPrevUB = PrevUB.get();
+ Expr *BoundPrevUB = NewPrevUB.get();
if (UseStrictCompare) {
BoundPrevUB =
SemaRef
diff --git a/clang/test/OpenMP/amdgcn_target_device_vla.cpp b/clang/test/OpenMP/amdgcn_target_device_vla.cpp
index 8b05f0aba04cec..0ec050879e4a27 100644
--- a/clang/test/OpenMP/amdgcn_target_device_vla.cpp
+++ b/clang/test/OpenMP/amdgcn_target_device_vla.cpp
@@ -454,9 +454,9 @@ int main() {
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK: omp.inner.for.cond:
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
-// CHECK-NEXT: [[CONV5:%.*]] = sext i32 [[TMP11]] to i64
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
-// CHECK-NEXT: [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP12]]
+// CHECK-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP12]] to i32
+// CHECK-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[CONV5]]
// CHECK-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK: omp.inner.for.body:
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
diff --git a/clang/test/OpenMP/amdgpu_target_with_aligned_attribute.c b/clang/test/OpenMP/amdgpu_target_with_aligned_attribute.c
index 1a641a118e2cdc..0fafc4f5f3b62a 100644
--- a/clang/test/OpenMP/amdgpu_target_with_aligned_attribute.c
+++ b/clang/test/OpenMP/amdgpu_target_with_aligned_attribute.c
@@ -270,9 +270,9 @@ void write_to_aligned_array(int *a, int N) {
// CHECK-AMD-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-AMD: omp.inner.for.cond:
// CHECK-AMD-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
-// CHECK-AMD-NEXT: [[CONV5:%.*]] = sext i32 [[TMP9]] to i64
// CHECK-AMD-NEXT: [[TMP10:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
-// CHECK-AMD-NEXT: [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP10]]
+// CHECK-AMD-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP10]] to i32
+// CHECK-AMD-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP9]], [[CONV5]]
// CHECK-AMD-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-AMD: omp.inner.for.body:
// CHECK-AMD-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
diff --git a/clang/test/OpenMP/distribute_parallel_for_bound_type_codegen.c b/clang/test/OpenMP/distribute_parallel_for_bound_type_codegen.c
new file mode 100644
index 00000000000000..428316b59786e7
--- /dev/null
+++ b/clang/test/OpenMP/distribute_parallel_for_bound_type_codegen.c
@@ -0,0 +1,30 @@
+// RUN: %clang_cc1 -fopenmp -triple x86_64-unknown-linux-gnu -emit-llvm -o - %s | FileCheck %s
+
+void signed_loop(float *V, int N) {
+#pragma omp teams distribute parallel for schedule(static, 1)
+ for (int i = 0; i < N; ++i)
+ ++V[i];
+}
+
+// CHECK-LABEL: define internal void @signed_loop.omp_outlined.omp_outlined(
+// CHECK: call void @__kmpc_for_static_init_4(
+// CHECK: omp.inner.for.cond:
+// CHECK-NEXT: [[IV:%.*]] = load i32, ptr %.omp.iv, align 4
+// CHECK-NEXT: [[PREV_UB:%.*]] = load i64, ptr %.previous.ub..addr, align 8
+// CHECK-NEXT: [[BOUND:%.*]] = trunc i64 [[PREV_UB]] to i32
+// CHECK-NEXT: [[CMP:%.*]] = icmp sle i32 [[IV]], [[BOUND]]
+
+void unsigned_loop(float *V, unsigned N) {
+#pragma omp teams distribute parallel for schedule(static, 1)
+ for (unsigned i = 0; i < N; ++i)
+ ++V[i];
+}
+
+// CHECK-LABEL: define internal void @unsigned_loop.omp_outlined.omp_outlined(
+// CHECK: call void @__kmpc_for_static_init_4u(
+// CHECK: omp.inner.for.cond:
+// CHECK-NEXT: [[IV:%.*]] = load i32, ptr %.omp.iv, align 4
+// CHECK-NEXT: [[PREV_UB:%.*]] = load i64, ptr %.previous.ub..addr, align 8
+// CHECK-NEXT: [[BOUND:%.*]] = trunc i64 [[PREV_UB]] to i32
+// CHECK-NEXT: [[NEXT_BOUND:%.*]] = add i32 [[BOUND]], 1
+// CHECK-NEXT: [[CMP:%.*]] = icmp ult i32 [[IV]], [[NEXT_BOUND]]
diff --git a/clang/test/OpenMP/nvptx_SPMD_codegen.cpp b/clang/test/OpenMP/nvptx_SPMD_codegen.cpp
index d15539a4712130..7295465178763e 100644
--- a/clang/test/OpenMP/nvptx_SPMD_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_SPMD_codegen.cpp
@@ -519,9 +519,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP72:![0-9]+]]
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-64-NEXT: [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP72]]
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP72]]
@@ -548,9 +548,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND4:%.*]]
// CHECK-64: omp.inner.for.cond4:
// CHECK-64-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK-64-NEXT: [[CONV5:%.*]] = sext i32 [[TMP14]] to i64
// CHECK-64-NEXT: [[TMP15:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK-64-NEXT: [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP15]]
+// CHECK-64-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP15]] to i32
+// CHECK-64-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[CONV5]]
// CHECK-64-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY7:%.*]], label [[OMP_INNER_FOR_END13:%.*]]
// CHECK-64: omp.inner.for.body7:
// CHECK-64-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -626,9 +626,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP76:![0-9]+]]
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
// CHECK-64-NEXT: [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP76]]
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP76]]
@@ -655,9 +655,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND4:%.*]]
// CHECK-64: omp.inner.for.cond4:
// CHECK-64-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK-64-NEXT: [[CONV5:%.*]] = sext i32 [[TMP14]] to i64
// CHECK-64-NEXT: [[TMP15:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK-64-NEXT: [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP15]]
+// CHECK-64-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP15]] to i32
+// CHECK-64-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[CONV5]]
// CHECK-64-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY7:%.*]], label [[OMP_INNER_FOR_END13:%.*]]
// CHECK-64: omp.inner.for.body7:
// CHECK-64-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -1036,9 +1036,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP89:![0-9]+]]
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP89]]
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP89]]
@@ -2009,9 +2009,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK-64-NEXT: [[CONV3:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV3]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2368,9 +2368,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -3258,9 +3258,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP131:![0-9]+]]
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP131]]
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP131]]
@@ -3621,9 +3621,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP143:![0-9]+]]
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP143]]
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP143]]
@@ -4586,9 +4586,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -4935,9 +4935,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -5817,9 +5817,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -6166,9 +6166,9 @@ int a;
// CHECK-64-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK-64: omp.inner.for.cond:
// CHECK-64-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK-64-NEXT: [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
// CHECK-64-NEXT: [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK-64-NEXT: [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP6]]
+// CHECK-64-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK-64-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[CONV2]]
// CHECK-64-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-64: omp.inner.for.body:
// CHECK-64-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -9520,7 +9520,7 @@ int a;
// CHECK-32: omp.inner.for.cond:
// CHECK-32-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP72:![0-9]+]]
// CHECK-32-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP72]]
-// CHECK-32-NEXT: [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK-32-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32: omp.inner.for.body:
// CHECK-32-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP72]]
@@ -9548,7 +9548,7 @@ int a;
// CHECK-32: omp.inner.for.cond2:
// CHECK-32-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-NEXT: [[CMP3:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY4:%.*]], label [[OMP_INNER_FOR_END10:%.*]]
// CHECK-32: omp.inner.for.body4:
// CHECK-32-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -9623,7 +9623,7 @@ int a;
// CHECK-32: omp.inner.for.cond:
// CHECK-32-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP76:![0-9]+]]
// CHECK-32-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP76]]
-// CHECK-32-NEXT: [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK-32-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK-32-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32: omp.inner.for.body:
// CHECK-32-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP76]]
@@ -9651,7 +9651,7 @@ int a;
// CHECK-32: omp.inner.for.cond2:
// CHECK-32-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-NEXT: [[CMP3:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
+// CHECK-32-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK-32-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY4:%.*]], label [[OMP_INNER_FOR_END10:%.*]]
// CHECK-32: omp.inner.for.body4:
// CHECK-32-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -10023,7 +10023,7 @@ int a;
// CHECK-32: omp.inner.for.cond:
// CHECK-32-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP89:![0-9]+]]
// CHECK-32-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP89]]
-// CHECK-32-NEXT: [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32: omp.inner.for.body:
// CHECK-32-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP89]]
@@ -10975,7 +10975,7 @@ int a;
// CHECK-32: omp.inner.for.cond:
// CHECK-32-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-NEXT: [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32: omp.inner.for.body:
// CHECK-32-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -11325,7 +11325,7 @@ int a;
// CHECK-32: omp.inner.for.cond:
// CHECK-32-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK-32-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-NEXT: [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32: omp.inner.for.body:
// CHECK-32-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -12194,7 +12194,7 @@ int a;
// CHECK-32: omp.inner.for.cond:
// CHECK-32-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP131:![0-9]+]]
// CHECK-32-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP131]]
-// CHECK-32-NEXT: [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK-32-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK-32: omp.inner.for.body:
// CHECK-32-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP131]]
@@ -12548,7 +12548,7 @@ int a;
// CHECK-32: omp.inner.for.cond:
// CHECK-32-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access....
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/239868
More information about the cfe-commits
mailing list