[clang] [Clang][OpenMP] Fix static,1 inner loop bound (PR #239868)

Amadeu Moya Sardà via cfe-commits cfe-commits at lists.llvm.org
Fri Nov 6 17:13:12 PST 2026


https://github.com/amad00r created https://github.com/llvm/llvm-project/pull/239868

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.

>From 0ea1c5d05d8e52b4b4bfe25136902bf96acb86d0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Amadeu=20Moya=20Sard=C3=A0?= <moyasarda1 at llnl.gov>
Date: Fri, 6 Nov 2026 17:01:02 -0800
Subject: [PATCH] [Clang][OpenMP] Fix the inner loop bound in static,1
 distribute parallel for

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.
---
 clang/lib/Sema/SemaOpenMP.cpp                 |   2 +-
 .../test/OpenMP/amdgcn_target_device_vla.cpp  |   4 +-
 .../amdgpu_target_with_aligned_attribute.c    |   4 +-
 ...stribute_parallel_for_bound_type_codegen.c |  30 +++++
 clang/test/OpenMP/nvptx_SPMD_codegen.cpp      | 104 +++++++++---------
 ...stribute_parallel_generic_mode_codegen.cpp |   6 +-
 ..._teams_distribute_parallel_for_codegen.cpp |  46 ++++----
 ...s_distribute_parallel_for_simd_codegen.cpp |  18 +--
 ...vptx_target_teams_generic_loop_codegen.cpp |  46 ++++----
 .../target_teams_generic_loop_codegen.cpp     |   4 +-
 ...s_generic_loop_codegen_as_parallel_for.cpp |  12 +-
 11 files changed, 153 insertions(+), 123 deletions(-)
 create mode 100644 clang/test/OpenMP/distribute_parallel_for_bound_type_codegen.c

diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index 6f4c71b118d59..baad51b6321f4 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 8b05f0aba04ce..0ec050879e4a2 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 1a641a118e2cd..0fafc4f5f3b62 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 0000000000000..428316b59786e
--- /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 d15539a471213..7295465178763 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.group [[ACC_GRP143:![0-9]+]]
 // CHECK-32-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP143]]
-// 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_GRP143]]
@@ -13492,7 +13492,7 @@ 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
@@ -13832,7 +13832,7 @@ 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
@@ -14693,7 +14693,7 @@ 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
@@ -15033,7 +15033,7 @@ 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
@@ -18369,7 +18369,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP72:![0-9]+]]
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP72]]
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP72]]
@@ -18397,7 +18397,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond2:
 // CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-EX-NEXT:    [[CMP3:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
+// CHECK-32-EX-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY4:%.*]], label [[OMP_INNER_FOR_END10:%.*]]
 // CHECK-32-EX:       omp.inner.for.body4:
 // CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -18472,7 +18472,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP76:![0-9]+]]
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP76]]
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP76]]
@@ -18500,7 +18500,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond2:
 // CHECK-32-EX-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK-32-EX-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-EX-NEXT:    [[CMP3:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
+// CHECK-32-EX-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY4:%.*]], label [[OMP_INNER_FOR_END10:%.*]]
 // CHECK-32-EX:       omp.inner.for.body4:
 // CHECK-32-EX-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -18872,7 +18872,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP89:![0-9]+]]
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP89]]
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP89]]
@@ -19824,7 +19824,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -20174,7 +20174,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -21043,7 +21043,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP131:![0-9]+]]
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP131]]
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP131]]
@@ -21397,7 +21397,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP143:![0-9]+]]
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP143]]
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP143]]
@@ -22341,7 +22341,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -22681,7 +22681,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -23542,7 +23542,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -23882,7 +23882,7 @@ int a;
 // CHECK-32-EX:       omp.inner.for.cond:
 // CHECK-32-EX-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK-32-EX-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP5]], [[TMP6]]
+// CHECK-32-EX-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK-32-EX-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK-32-EX:       omp.inner.for.body:
 // CHECK-32-EX-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
diff --git a/clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp b/clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
index 9f5d142637c68..e8fb6153a8aa6 100644
--- a/clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_distribute_parallel_generic_mode_codegen.cpp
@@ -287,9 +287,9 @@ int main(int argc, char **argv) {
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:
 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK4-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP14]] to i64
 // CHECK4-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK4-NEXT:    [[CMP8:%.*]] = icmp ule i64 [[CONV7]], [[TMP15]]
+// CHECK4-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP15]] to i32
+// CHECK4-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP14]], [[CONV7]]
 // CHECK4-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK4:       omp.inner.for.body:
 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -602,7 +602,7 @@ int main(int argc, char **argv) {
 // CHECK5:       omp.inner.for.cond:
 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK5-NEXT:    [[CMP6:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
+// CHECK5-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK5-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
diff --git a/clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp b/clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp
index 8ebe1fcad87ab..61f85210755a8 100644
--- a/clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp
@@ -602,9 +602,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK1-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -795,9 +795,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
+// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV2]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -1005,9 +1005,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
+// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV3]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -1303,7 +1303,7 @@ int bar(int n){
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP11:%.*]] = icmp ule i64 [[TMP13]], [[TMP14]]
+// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i64 [[TMP13]], [[TMP14]]
 // CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
@@ -1593,9 +1593,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK1-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2162,9 +2162,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK2-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2355,9 +2355,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
+// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV2]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2565,9 +2565,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
+// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV3]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2861,9 +2861,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV11:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP12:%.*]] = icmp ule i64 [[CONV11]], [[TMP14]]
+// CHECK2-NEXT:    [[CONV11:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK2-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP13]], [[CONV11]]
 // CHECK2-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -3148,9 +3148,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK2-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -3707,7 +3707,7 @@ int bar(int n){
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP4:%.*]] = icmp ule i32 [[TMP10]], [[TMP11]]
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -3894,7 +3894,7 @@ int bar(int n){
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -4098,7 +4098,7 @@ int bar(int n){
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -4682,7 +4682,7 @@ int bar(int n){
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP4:%.*]] = icmp ule i32 [[TMP10]], [[TMP11]]
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
diff --git a/clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp b/clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp
index 1dfb22fac6f3b..3a2ecad5c24de 100644
--- a/clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp
@@ -626,9 +626,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20:![0-9]+]]
-// CHECK1-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP20]]
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK1-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]]
@@ -838,9 +838,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP26:![0-9]+]]
-// CHECK1-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP26]]
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
+// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV2]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP26]]
@@ -1063,9 +1063,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP32:![0-9]+]]
-// CHECK1-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group [[ACC_GRP32]]
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
+// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV3]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP32]]
@@ -1677,7 +1677,7 @@ int bar(int n){
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20:![0-9]+]]
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP20]]
-// CHECK2-NEXT:    [[CMP4:%.*]] = icmp ule i32 [[TMP10]], [[TMP11]]
+// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]]
@@ -1883,7 +1883,7 @@ int bar(int n){
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP26:![0-9]+]]
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP26]]
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP26]]
@@ -2102,7 +2102,7 @@ int bar(int n){
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP32:![0-9]+]]
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group [[ACC_GRP32]]
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP32]]
diff --git a/clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp b/clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp
index c97fa6b299688..a5d032cf5171c 100644
--- a/clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp
@@ -282,9 +282,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK1-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -485,9 +485,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
+// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV2]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -695,9 +695,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
+// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV3]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -993,7 +993,7 @@ int bar(int n){
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP11:%.*]] = icmp ule i64 [[TMP13]], [[TMP14]]
+// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i64 [[TMP13]], [[TMP14]]
 // CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
@@ -1283,9 +1283,9 @@ int bar(int n){
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK1-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -1538,9 +1538,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK2-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -1741,9 +1741,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
+// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV2]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -1951,9 +1951,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV3]], [[TMP7]]
+// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[CONV3]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2247,9 +2247,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV11:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP12:%.*]] = icmp ule i64 [[CONV11]], [[TMP14]]
+// CHECK2-NEXT:    [[CONV11:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK2-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP13]], [[CONV11]]
 // CHECK2-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2534,9 +2534,9 @@ int bar(int n){
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK2-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[CONV5]]
 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2786,7 +2786,7 @@ int bar(int n){
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP4:%.*]] = icmp ule i32 [[TMP10]], [[TMP11]]
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -2983,7 +2983,7 @@ int bar(int n){
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -3187,7 +3187,7 @@ int bar(int n){
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp ule i32 [[TMP6]], [[TMP7]]
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
@@ -3771,7 +3771,7 @@ int bar(int n){
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP4:%.*]] = icmp ule i32 [[TMP10]], [[TMP11]]
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
diff --git a/clang/test/OpenMP/target_teams_generic_loop_codegen.cpp b/clang/test/OpenMP/target_teams_generic_loop_codegen.cpp
index 4b5eaee76c06c..f7619ac90011f 100644
--- a/clang/test/OpenMP/target_teams_generic_loop_codegen.cpp
+++ b/clang/test/OpenMP/target_teams_generic_loop_codegen.cpp
@@ -255,9 +255,9 @@ int foo() {
 // IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // IR-GPU:       omp.inner.for.cond:
 // IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7:![0-9]+]]
-// IR-GPU-NEXT:    [[CONV6:%.*]] = sext i32 [[TMP7]] to i64
 // IR-GPU-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8, !llvm.access.group [[ACC_GRP7]]
-// IR-GPU-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV6]], [[TMP8]]
+// IR-GPU-NEXT:    [[CONV6:%.*]] = trunc i64 [[TMP8]] to i32
+// IR-GPU-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP7]], [[CONV6]]
 // IR-GPU-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // IR-GPU:       omp.inner.for.body:
 // IR-GPU-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
diff --git a/clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp b/clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp
index a16e3942526c5..9c37e30f72d87 100644
--- a/clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp
+++ b/clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp
@@ -368,9 +368,9 @@ int main()
 // IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // IR-GPU:       omp.inner.for.cond:
 // IR-GPU-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
-// IR-GPU-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // IR-GPU-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
-// IR-GPU-NEXT:    [[CMP8:%.*]] = icmp ule i64 [[CONV7]], [[TMP14]]
+// IR-GPU-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// IR-GPU-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP13]], [[CONV7]]
 // IR-GPU-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // IR-GPU:       omp.inner.for.body:
 // IR-GPU-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
@@ -742,7 +742,7 @@ int main()
 // IR-GPU:       omp.inner.for.cond:
 // IR-GPU-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
 // IR-GPU-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
-// IR-GPU-NEXT:    [[CMP13:%.*]] = icmp ule i64 [[TMP16]], [[TMP17]]
+// IR-GPU-NEXT:    [[CMP13:%.*]] = icmp sle i64 [[TMP16]], [[TMP17]]
 // IR-GPU-NEXT:    br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // IR-GPU:       omp.inner.for.body:
 // IR-GPU-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
@@ -1126,9 +1126,9 @@ int main()
 // IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // IR-GPU:       omp.inner.for.cond:
 // IR-GPU-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
-// IR-GPU-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // IR-GPU-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
-// IR-GPU-NEXT:    [[CMP8:%.*]] = icmp ule i64 [[CONV7]], [[TMP14]]
+// IR-GPU-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// IR-GPU-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP13]], [[CONV7]]
 // IR-GPU-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // IR-GPU:       omp.inner.for.body:
 // IR-GPU-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
@@ -3273,7 +3273,7 @@ int main()
 // IR-GPU-NESTED:       omp.inner.for.cond:
 // IR-GPU-NESTED-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
 // IR-GPU-NESTED-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
-// IR-GPU-NESTED-NEXT:    [[CMP13:%.*]] = icmp ule i64 [[TMP16]], [[TMP17]]
+// IR-GPU-NESTED-NEXT:    [[CMP13:%.*]] = icmp sle i64 [[TMP16]], [[TMP17]]
 // IR-GPU-NESTED-NEXT:    br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // IR-GPU-NESTED:       omp.inner.for.body:
 // IR-GPU-NESTED-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8



More information about the cfe-commits mailing list