[OPENMP]Fix PR50640: OpenMP target clause implicitly scaling loop bounds to uint64_t.

[ABataev]

Summary

Need to add some conversions to suppress possible warning messages.

[jhuber6]

LGTM, work on my end.

clang/lib/Sema/SemaOpenMP.cpp

if (isOpenMPLoopBoundSharingDirective(DKind)) {
     DistInc = SemaRef.BuildBinOp(CurScope, DistIncLoc, BO_Assign, IV.get(),
                                  DistInc.get());
     DistInc =
         SemaRef.ActOnFinishFullExpr(DistInc.get(), /*DiscardedValue*/ false);
     assert(DistInc.isUsable() && "distribute inc expr was not built");
 
     // Build expression: UB = min(UB, prevUB) for #for in composite or combined
     // construct
+    ExprResult NewPrevUB = PrevUB;
+    if (!SemaRef.Context.hasSameType(UB.get()->getType(),
+                                     PrevUB.get()->getType())) {
+      NewPrevUB = SemaRef.PerformImplicitConversion(
+          NewPrevUB.get(), UB.get()->getType(), Sema::AA_Converting,
+          /*AllowExplicit=*/true);
+      if (!NewPrevUB.isUsable())
+        return 0;
+    }
     SourceLocation DistEUBLoc = AStmt->getBeginLoc();
-    ExprResult IsUBGreater =
-        SemaRef.BuildBinOp(CurScope, DistEUBLoc, BO_GT, UB.get(), PrevUB.get());
+    ExprResult IsUBGreater = SemaRef.BuildBinOp(CurScope, DistEUBLoc, BO_GT,
+                                                UB.get(), NewPrevUB.get());
     ExprResult CondOp = SemaRef.ActOnConditionalOp(
-        DistEUBLoc, DistEUBLoc, IsUBGreater.get(), PrevUB.get(), UB.get());
+        DistEUBLoc, DistEUBLoc, IsUBGreater.get(), NewPrevUB.get(), UB.get());
     PrevEUB = SemaRef.BuildBinOp(CurScope, DistIncLoc, BO_Assign, UB.get(),
                                  CondOp.get());
     PrevEUB =
         SemaRef.ActOnFinishFullExpr(PrevEUB.get(), /*DiscardedValue*/ false);
 
     // 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)

clang/test/OpenMP/distribute_parallel_for_codegen.cpp

 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK1-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK1-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK1-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK1-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK1-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
 // CHECK1-NEXT:    [[TMP23:%.*]] = load double*, double** [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 [[IDXPROM]]
 // CHECK1-NEXT:    [[TMP25:%.*]] = load double, double* [[ARRAYIDX]], align 8
 // CHECK1-NEXT:    [[TMP26:%.*]] = load double*, double** [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK1-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK1-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX14]], align 8
-// CHECK1-NEXT:    [[ADD15:%.*]] = fadd double [[TMP25]], [[TMP28]]
+// CHECK1-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK1-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX13]], align 8
+// CHECK1-NEXT:    [[ADD14:%.*]] = fadd double [[TMP25]], [[TMP28]]
 // CHECK1-NEXT:    [[TMP29:%.*]] = load double*, double** [[TMP1]], align 8
 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK1-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK1-NEXT:    store double [[ADD15]], double* [[ARRAYIDX17]], align 8
+// CHECK1-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK1-NEXT:    store double [[ADD14]], double* [[ARRAYIDX16]], align 8
 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 0
 // CHECK1-NEXT:    store double** [[TMP1]], double*** [[TMP31]], align 8
 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 1
 // CHECK1-NEXT:    store i32* [[I6]], i32** [[TMP32]], align 8
 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 2
 // CHECK1-NEXT:    store double** [[TMP2]], double*** [[TMP33]], align 8
 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 3
 // CHECK1-NEXT:    store double** [[TMP3]], double*** [[TMP34]], align 8
 // CHECK1-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE3_clEv"(%class.anon.4* nonnull align 8 dereferenceable(32) [[REF_TMP]])
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP35]], 1
-// CHECK1-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP35]], 1
+// CHECK1-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
-// CHECK1-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
+// CHECK1-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
-// CHECK1-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
+// CHECK1-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP41]])
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void

 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK2-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK2-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK2-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK2-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
-// CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK2-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK2-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK2-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK2-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK2-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK2-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
 // CHECK2-NEXT:    [[TMP23:%.*]] = load double*, double** [[TMP2]], align 8
 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 [[IDXPROM]]
 // CHECK2-NEXT:    [[TMP25:%.*]] = load double, double* [[ARRAYIDX]], align 8
 // CHECK2-NEXT:    [[TMP26:%.*]] = load double*, double** [[TMP3]], align 8
 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK2-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK2-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK2-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX14]], align 8
-// CHECK2-NEXT:    [[ADD15:%.*]] = fadd double [[TMP25]], [[TMP28]]
+// CHECK2-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK2-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK2-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX13]], align 8
+// CHECK2-NEXT:    [[ADD14:%.*]] = fadd double [[TMP25]], [[TMP28]]
 // CHECK2-NEXT:    [[TMP29:%.*]] = load double*, double** [[TMP1]], align 8
 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK2-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK2-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK2-NEXT:    store double [[ADD15]], double* [[ARRAYIDX17]], align 8
+// CHECK2-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK2-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK2-NEXT:    store double [[ADD14]], double* [[ARRAYIDX16]], align 8
 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 0
 // CHECK2-NEXT:    store double** [[TMP1]], double*** [[TMP31]], align 8
 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 1
 // CHECK2-NEXT:    store i32* [[I6]], i32** [[TMP32]], align 8
 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 2
 // CHECK2-NEXT:    store double** [[TMP2]], double*** [[TMP33]], align 8
 // CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 3
 // CHECK2-NEXT:    store double** [[TMP3]], double*** [[TMP34]], align 8
 // CHECK2-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE3_clEv"(%class.anon.4* nonnull align 8 dereferenceable(32) [[REF_TMP]])
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP35]], 1
-// CHECK2-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP35]], 1
+// CHECK2-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
-// CHECK2-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
+// CHECK2-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
-// CHECK2-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
+// CHECK2-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP41]])
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       omp.precond.end:
 // CHECK2-NEXT:    ret void

 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:

 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK4-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK4-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:

 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK9:       omp.dispatch.cond:
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK9-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK9-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK9-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK9-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK9-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK9-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK9:       omp.dispatch.body:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK9-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK9-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
 // CHECK9-NEXT:    [[TMP23:%.*]] = load double*, double** [[TMP2]], align 8
 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 [[IDXPROM]]
 // CHECK9-NEXT:    [[TMP25:%.*]] = load double, double* [[ARRAYIDX]], align 8
 // CHECK9-NEXT:    [[TMP26:%.*]] = load double*, double** [[TMP3]], align 8
 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK9-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK9-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX14]], align 8
-// CHECK9-NEXT:    [[ADD15:%.*]] = fadd double [[TMP25]], [[TMP28]]
+// CHECK9-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK9-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX13]], align 8
+// CHECK9-NEXT:    [[ADD14:%.*]] = fadd double [[TMP25]], [[TMP28]]
 // CHECK9-NEXT:    [[TMP29:%.*]] = load double*, double** [[TMP1]], align 8
 // CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK9-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK9-NEXT:    store double [[ADD15]], double* [[ARRAYIDX17]], align 8
+// CHECK9-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK9-NEXT:    store double [[ADD14]], double* [[ARRAYIDX16]], align 8
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP31]], 1
-// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP31]], 1
+// CHECK9-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK9:       omp.dispatch.inc:
 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
-// CHECK9-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
+// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
-// CHECK9-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
+// CHECK9-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK9:       omp.dispatch.end:
 // CHECK9-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP37]])
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void

 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK9:       omp.dispatch.cond:
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK9-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK9-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK9-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK9-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK9-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK9-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK9:       omp.dispatch.body:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK9-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK9-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[TMP2]], align 8
 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i64 [[IDXPROM]]
 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK9-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP3]], align 8
 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK9-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX14]], align 4
-// CHECK9-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP25]], [[TMP28]]
+// CHECK9-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX13]], align 4
+// CHECK9-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP25]], [[TMP28]]
 // CHECK9-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[TMP1]], align 8
 // CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK9-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK9-NEXT:    store i32 [[ADD15]], i32* [[ARRAYIDX17]], align 4
+// CHECK9-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK9-NEXT:    store i32 [[ADD14]], i32* [[ARRAYIDX16]], align 4
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP31]], 1
-// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP31]], 1
+// CHECK9-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK9:       omp.dispatch.inc:
 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
-// CHECK9-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
+// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
-// CHECK9-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
+// CHECK9-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK9:       omp.dispatch.end:
 // CHECK9-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP37]])
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void

 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK10:       omp.dispatch.cond:
 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK10-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK10-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK10-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK10-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK10-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK10:       cond.true:
 // CHECK10-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK10-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK10-NEXT:    br label [[COND_END:%.*]]
 // CHECK10:       cond.false:
 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK10-NEXT:    br label [[COND_END]]
 // CHECK10:       cond.end:
-// CHECK10-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK10-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK10-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK10-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK10-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK10:       omp.dispatch.body:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK10:       omp.inner.for.cond:
 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK10-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK10-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK10:       omp.inner.for.body:
 // CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
 // CHECK10-NEXT:    [[TMP23:%.*]] = load double*, double** [[TMP2]], align 8
 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 [[IDXPROM]]
 // CHECK10-NEXT:    [[TMP25:%.*]] = load double, double* [[ARRAYIDX]], align 8
 // CHECK10-NEXT:    [[TMP26:%.*]] = load double*, double** [[TMP3]], align 8
 // CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK10-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK10-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK10-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX14]], align 8
-// CHECK10-NEXT:    [[ADD15:%.*]] = fadd double [[TMP25]], [[TMP28]]
+// CHECK10-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK10-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK10-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX13]], align 8
+// CHECK10-NEXT:    [[ADD14:%.*]] = fadd double [[TMP25]], [[TMP28]]
 // CHECK10-NEXT:    [[TMP29:%.*]] = load double*, double** [[TMP1]], align 8
 // CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK10-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK10-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK10-NEXT:    store double [[ADD15]], double* [[ARRAYIDX17]], align 8
+// CHECK10-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK10-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK10-NEXT:    store double [[ADD14]], double* [[ARRAYIDX16]], align 8
 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK10:       omp.body.continue:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK10:       omp.inner.for.inc:
 // CHECK10-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP31]], 1
-// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4
+// CHECK10-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP31]], 1
+// CHECK10-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK10:       omp.inner.for.end:
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK10:       omp.dispatch.inc:
 // CHECK10-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK10-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
-// CHECK10-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
+// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK10-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
-// CHECK10-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
+// CHECK10-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK10:       omp.dispatch.end:
 // CHECK10-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP37]])
 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK10:       omp.precond.end:
 // CHECK10-NEXT:    ret void

 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK10:       omp.dispatch.cond:
 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK10-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK10-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK10-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK10-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK10-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK10:       cond.true:
 // CHECK10-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK10-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK10-NEXT:    br label [[COND_END:%.*]]
 // CHECK10:       cond.false:
 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK10-NEXT:    br label [[COND_END]]
 // CHECK10:       cond.end:
-// CHECK10-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK10-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK10-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK10-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK10-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK10:       omp.dispatch.body:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK10:       omp.inner.for.cond:
 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK10-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK10-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK10:       omp.inner.for.body:
 // CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[TMP2]], align 8
 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i64 [[IDXPROM]]
 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK10-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP3]], align 8
 // CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK10-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK10-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX14]], align 4
-// CHECK10-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP25]], [[TMP28]]
+// CHECK10-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK10-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX13]], align 4
+// CHECK10-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP25]], [[TMP28]]
 // CHECK10-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[TMP1]], align 8
 // CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK10-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK10-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK10-NEXT:    store i32 [[ADD15]], i32* [[ARRAYIDX17]], align 4
+// CHECK10-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK10-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK10-NEXT:    store i32 [[ADD14]], i32* [[ARRAYIDX16]], align 4
 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK10:       omp.body.continue:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK10:       omp.inner.for.inc:
 // CHECK10-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP31]], 1
-// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4
+// CHECK10-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP31]], 1
+// CHECK10-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK10:       omp.inner.for.end:
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK10:       omp.dispatch.inc:
 // CHECK10-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK10-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
-// CHECK10-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
+// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK10-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
-// CHECK10-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
+// CHECK10-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK10:       omp.dispatch.end:
 // CHECK10-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP37]])
 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK10:       omp.precond.end:
 // CHECK10-NEXT:    ret void

 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK11:       omp.dispatch.cond:
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:

 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK11:       omp.dispatch.cond:
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:

 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK12-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK12:       omp.dispatch.cond:
 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK12-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK12-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK12-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK12:       cond.true:
 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK12-NEXT:    br label [[COND_END:%.*]]
 // CHECK12:       cond.false:
 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK12-NEXT:    br label [[COND_END]]
 // CHECK12:       cond.end:

 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK12-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK12:       omp.dispatch.cond:
 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK12-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK12-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK12-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK12:       cond.true:
 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK12-NEXT:    br label [[COND_END:%.*]]
 // CHECK12:       cond.false:
 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK12-NEXT:    br label [[COND_END]]
 // CHECK12:       cond.end:

clang/test/OpenMP/distribute_parallel_for_simd_codegen.cpp

 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK1-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK1-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK1-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK1-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !40
-// CHECK1-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK1-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP23:%.*]] = load double*, double** [[TMP2]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !40
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 [[IDXPROM]]
 // CHECK1-NEXT:    [[TMP25:%.*]] = load double, double* [[ARRAYIDX]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP26:%.*]] = load double*, double** [[TMP3]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !40
-// CHECK1-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK1-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX14]], align 8, !llvm.access.group !40
-// CHECK1-NEXT:    [[ADD15:%.*]] = fadd double [[TMP25]], [[TMP28]]
+// CHECK1-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK1-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX13]], align 8, !llvm.access.group !40
+// CHECK1-NEXT:    [[ADD14:%.*]] = fadd double [[TMP25]], [[TMP28]]
 // CHECK1-NEXT:    [[TMP29:%.*]] = load double*, double** [[TMP1]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !40
-// CHECK1-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK1-NEXT:    store double [[ADD15]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !40
+// CHECK1-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK1-NEXT:    store double [[ADD14]], double* [[ARRAYIDX16]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 0
 // CHECK1-NEXT:    store double** [[TMP1]], double*** [[TMP31]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 1
 // CHECK1-NEXT:    store i32* [[I6]], i32** [[TMP32]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 2
 // CHECK1-NEXT:    store double** [[TMP2]], double*** [[TMP33]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 3
 // CHECK1-NEXT:    store double** [[TMP3]], double*** [[TMP34]], align 8, !llvm.access.group !40
 // CHECK1-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE3_clEv"(%class.anon.4* nonnull align 8 dereferenceable(32) [[REF_TMP]]), !llvm.access.group !40
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
-// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP35]], 1
-// CHECK1-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
+// CHECK1-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP35]], 1
+// CHECK1-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
-// CHECK1-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
+// CHECK1-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
-// CHECK1-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
+// CHECK1-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP41]])
 // CHECK1-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
 // CHECK1-NEXT:    br i1 [[TMP43]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    [[SUB21:%.*]] = sub nsw i32 [[TMP44]], 0
-// CHECK1-NEXT:    [[DIV22:%.*]] = sdiv i32 [[SUB21]], 1
-// CHECK1-NEXT:    [[MUL23:%.*]] = mul nsw i32 [[DIV22]], 1
-// CHECK1-NEXT:    [[ADD24:%.*]] = add nsw i32 0, [[MUL23]]
-// CHECK1-NEXT:    store i32 [[ADD24]], i32* [[I6]], align 4
+// CHECK1-NEXT:    [[SUB20:%.*]] = sub nsw i32 [[TMP44]], 0
+// CHECK1-NEXT:    [[DIV21:%.*]] = sdiv i32 [[SUB20]], 1
+// CHECK1-NEXT:    [[MUL22:%.*]] = mul nsw i32 [[DIV21]], 1
+// CHECK1-NEXT:    [[ADD23:%.*]] = add nsw i32 0, [[MUL22]]
+// CHECK1-NEXT:    store i32 [[ADD23]], i32* [[I6]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l299

 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK2-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK2-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK2-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK2-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
-// CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK2-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK2-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK2-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK2-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !40
-// CHECK2-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK2-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK2-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP23:%.*]] = load double*, double** [[TMP2]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !40
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 [[IDXPROM]]
 // CHECK2-NEXT:    [[TMP25:%.*]] = load double, double* [[ARRAYIDX]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP26:%.*]] = load double*, double** [[TMP3]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !40
-// CHECK2-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK2-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK2-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX14]], align 8, !llvm.access.group !40
-// CHECK2-NEXT:    [[ADD15:%.*]] = fadd double [[TMP25]], [[TMP28]]
+// CHECK2-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK2-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK2-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX13]], align 8, !llvm.access.group !40
+// CHECK2-NEXT:    [[ADD14:%.*]] = fadd double [[TMP25]], [[TMP28]]
 // CHECK2-NEXT:    [[TMP29:%.*]] = load double*, double** [[TMP1]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !40
-// CHECK2-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK2-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK2-NEXT:    store double [[ADD15]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !40
+// CHECK2-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK2-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK2-NEXT:    store double [[ADD14]], double* [[ARRAYIDX16]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 0
 // CHECK2-NEXT:    store double** [[TMP1]], double*** [[TMP31]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 1
 // CHECK2-NEXT:    store i32* [[I6]], i32** [[TMP32]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 2
 // CHECK2-NEXT:    store double** [[TMP2]], double*** [[TMP33]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[CLASS_ANON_4]], %class.anon.4* [[REF_TMP]], i32 0, i32 3
 // CHECK2-NEXT:    store double** [[TMP3]], double*** [[TMP34]], align 8, !llvm.access.group !40
 // CHECK2-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE3_clEv"(%class.anon.4* nonnull align 8 dereferenceable(32) [[REF_TMP]]), !llvm.access.group !40
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
-// CHECK2-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP35]], 1
-// CHECK2-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
+// CHECK2-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP35]], 1
+// CHECK2-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
-// CHECK2-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
+// CHECK2-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
-// CHECK2-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
+// CHECK2-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP41]])
 // CHECK2-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
 // CHECK2-NEXT:    br i1 [[TMP43]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK2:       .omp.final.then:
 // CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK2-NEXT:    [[SUB21:%.*]] = sub nsw i32 [[TMP44]], 0
-// CHECK2-NEXT:    [[DIV22:%.*]] = sdiv i32 [[SUB21]], 1
-// CHECK2-NEXT:    [[MUL23:%.*]] = mul nsw i32 [[DIV22]], 1
-// CHECK2-NEXT:    [[ADD24:%.*]] = add nsw i32 0, [[MUL23]]
-// CHECK2-NEXT:    store i32 [[ADD24]], i32* [[I6]], align 4
+// CHECK2-NEXT:    [[SUB20:%.*]] = sub nsw i32 [[TMP44]], 0
+// CHECK2-NEXT:    [[DIV21:%.*]] = sdiv i32 [[SUB20]], 1
+// CHECK2-NEXT:    [[MUL22:%.*]] = mul nsw i32 [[DIV21]], 1
+// CHECK2-NEXT:    [[ADD23:%.*]] = add nsw i32 0, [[MUL22]]
+// CHECK2-NEXT:    store i32 [[ADD23]], i32* [[I6]], align 4
 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK2:       .omp.final.done:
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       omp.precond.end:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l299

 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:

 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK4-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK4-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:

 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK9:       omp.dispatch.cond:
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK9-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK9-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK9-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK9-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK9-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK9-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK9:       omp.dispatch.body:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !47
-// CHECK9-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK9-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK9-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !47
 // CHECK9-NEXT:    [[TMP23:%.*]] = load double*, double** [[TMP2]], align 8, !llvm.access.group !47
 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !47
 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 [[IDXPROM]]
 // CHECK9-NEXT:    [[TMP25:%.*]] = load double, double* [[ARRAYIDX]], align 8, !llvm.access.group !47
 // CHECK9-NEXT:    [[TMP26:%.*]] = load double*, double** [[TMP3]], align 8, !llvm.access.group !47
 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !47
-// CHECK9-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK9-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX14]], align 8, !llvm.access.group !47
-// CHECK9-NEXT:    [[ADD15:%.*]] = fadd double [[TMP25]], [[TMP28]]
+// CHECK9-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK9-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX13]], align 8, !llvm.access.group !47
+// CHECK9-NEXT:    [[ADD14:%.*]] = fadd double [[TMP25]], [[TMP28]]
 // CHECK9-NEXT:    [[TMP29:%.*]] = load double*, double** [[TMP1]], align 8, !llvm.access.group !47
 // CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !47
-// CHECK9-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK9-NEXT:    store double [[ADD15]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !47
+// CHECK9-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK9-NEXT:    store double [[ADD14]], double* [[ARRAYIDX16]], align 8, !llvm.access.group !47
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
-// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP31]], 1
-// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
+// CHECK9-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP31]], 1
+// CHECK9-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK9:       omp.dispatch.inc:
 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
-// CHECK9-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
+// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
-// CHECK9-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
+// CHECK9-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK9:       omp.dispatch.end:
 // CHECK9-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP37]])
 // CHECK9-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 // CHECK9-NEXT:    br i1 [[TMP39]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK9-NEXT:    [[SUB21:%.*]] = sub nsw i32 [[TMP40]], 0
-// CHECK9-NEXT:    [[DIV22:%.*]] = sdiv i32 [[SUB21]], 1
-// CHECK9-NEXT:    [[MUL23:%.*]] = mul nsw i32 [[DIV22]], 1
-// CHECK9-NEXT:    [[ADD24:%.*]] = add nsw i32 0, [[MUL23]]
-// CHECK9-NEXT:    store i32 [[ADD24]], i32* [[I6]], align 4
+// CHECK9-NEXT:    [[SUB20:%.*]] = sub nsw i32 [[TMP40]], 0
+// CHECK9-NEXT:    [[DIV21:%.*]] = sdiv i32 [[SUB20]], 1
+// CHECK9-NEXT:    [[MUL22:%.*]] = mul nsw i32 [[DIV21]], 1
+// CHECK9-NEXT:    [[ADD23:%.*]] = add nsw i32 0, [[MUL22]]
+// CHECK9-NEXT:    store i32 [[ADD23]], i32* [[I6]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l535

 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK9:       omp.dispatch.cond:
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK9-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK9-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK9-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK9-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK9-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK9-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK9:       omp.dispatch.body:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !89
-// CHECK9-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK9-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK9-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !89
 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[TMP2]], align 8, !llvm.access.group !89
 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !89
 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i64 [[IDXPROM]]
 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !89
 // CHECK9-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP3]], align 8, !llvm.access.group !89
 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !89
-// CHECK9-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX14]], align 4, !llvm.access.group !89
-// CHECK9-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP25]], [[TMP28]]
+// CHECK9-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX13]], align 4, !llvm.access.group !89
+// CHECK9-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP25]], [[TMP28]]
 // CHECK9-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[TMP1]], align 8, !llvm.access.group !89
 // CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !89
-// CHECK9-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK9-NEXT:    store i32 [[ADD15]], i32* [[ARRAYIDX17]], align 4, !llvm.access.group !89
+// CHECK9-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK9-NEXT:    store i32 [[ADD14]], i32* [[ARRAYIDX16]], align 4, !llvm.access.group !89
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
-// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP31]], 1
-// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
+// CHECK9-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP31]], 1
+// CHECK9-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP90:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK9:       omp.dispatch.inc:
 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
-// CHECK9-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
+// CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
-// CHECK9-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
+// CHECK9-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK9:       omp.dispatch.end:
 // CHECK9-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP37]])
 // CHECK9-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 // CHECK9-NEXT:    br i1 [[TMP39]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK9-NEXT:    [[SUB21:%.*]] = sub nsw i32 [[TMP40]], 0
-// CHECK9-NEXT:    [[DIV22:%.*]] = sdiv i32 [[SUB21]], 1
-// CHECK9-NEXT:    [[MUL23:%.*]] = mul nsw i32 [[DIV22]], 1
-// CHECK9-NEXT:    [[ADD24:%.*]] = add nsw i32 0, [[MUL23]]
-// CHECK9-NEXT:    store i32 [[ADD24]], i32* [[I6]], align 4
+// CHECK9-NEXT:    [[SUB20:%.*]] = sub nsw i32 [[TMP40]], 0
+// CHECK9-NEXT:    [[DIV21:%.*]] = sdiv i32 [[SUB20]], 1
+// CHECK9-NEXT:    [[MUL22:%.*]] = mul nsw i32 [[DIV21]], 1
+// CHECK9-NEXT:    [[ADD23:%.*]] = add nsw i32 0, [[MUL22]]
+// CHECK9-NEXT:    store i32 [[ADD23]], i32* [[I6]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l82

 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK10:       omp.dispatch.cond:
 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK10-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK10-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK10-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK10-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK10-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK10:       cond.true:
 // CHECK10-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK10-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK10-NEXT:    br label [[COND_END:%.*]]
 // CHECK10:       cond.false:
 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK10-NEXT:    br label [[COND_END]]
 // CHECK10:       cond.end:
-// CHECK10-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK10-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK10-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK10-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK10-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK10:       omp.dispatch.body:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK10:       omp.inner.for.cond:
 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !47
-// CHECK10-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK10-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK10-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK10:       omp.inner.for.body:
 // CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !47
 // CHECK10-NEXT:    [[TMP23:%.*]] = load double*, double** [[TMP2]], align 8, !llvm.access.group !47
 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !47
 // CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 [[IDXPROM]]
 // CHECK10-NEXT:    [[TMP25:%.*]] = load double, double* [[ARRAYIDX]], align 8, !llvm.access.group !47
 // CHECK10-NEXT:    [[TMP26:%.*]] = load double*, double** [[TMP3]], align 8, !llvm.access.group !47
 // CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !47
-// CHECK10-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK10-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK10-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX14]], align 8, !llvm.access.group !47
-// CHECK10-NEXT:    [[ADD15:%.*]] = fadd double [[TMP25]], [[TMP28]]
+// CHECK10-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK10-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK10-NEXT:    [[TMP28:%.*]] = load double, double* [[ARRAYIDX13]], align 8, !llvm.access.group !47
+// CHECK10-NEXT:    [[ADD14:%.*]] = fadd double [[TMP25]], [[TMP28]]
 // CHECK10-NEXT:    [[TMP29:%.*]] = load double*, double** [[TMP1]], align 8, !llvm.access.group !47
 // CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !47
-// CHECK10-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK10-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK10-NEXT:    store double [[ADD15]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !47
+// CHECK10-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK10-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK10-NEXT:    store double [[ADD14]], double* [[ARRAYIDX16]], align 8, !llvm.access.group !47
 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK10:       omp.body.continue:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK10:       omp.inner.for.inc:
 // CHECK10-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
-// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP31]], 1
-// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
+// CHECK10-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP31]], 1
+// CHECK10-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
 // CHECK10:       omp.inner.for.end:
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK10:       omp.dispatch.inc:
 // CHECK10-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK10-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
-// CHECK10-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
+// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK10-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
-// CHECK10-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
+// CHECK10-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK10:       omp.dispatch.end:
 // CHECK10-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP37]])
 // CHECK10-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK10-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 // CHECK10-NEXT:    br i1 [[TMP39]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK10:       .omp.final.then:
 // CHECK10-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK10-NEXT:    [[SUB21:%.*]] = sub nsw i32 [[TMP40]], 0
-// CHECK10-NEXT:    [[DIV22:%.*]] = sdiv i32 [[SUB21]], 1
-// CHECK10-NEXT:    [[MUL23:%.*]] = mul nsw i32 [[DIV22]], 1
-// CHECK10-NEXT:    [[ADD24:%.*]] = add nsw i32 0, [[MUL23]]
-// CHECK10-NEXT:    store i32 [[ADD24]], i32* [[I6]], align 4
+// CHECK10-NEXT:    [[SUB20:%.*]] = sub nsw i32 [[TMP40]], 0
+// CHECK10-NEXT:    [[DIV21:%.*]] = sdiv i32 [[SUB20]], 1
+// CHECK10-NEXT:    [[MUL22:%.*]] = mul nsw i32 [[DIV21]], 1
+// CHECK10-NEXT:    [[ADD23:%.*]] = add nsw i32 0, [[MUL22]]
+// CHECK10-NEXT:    store i32 [[ADD23]], i32* [[I6]], align 4
 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK10:       .omp.final.done:
 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK10:       omp.precond.end:
 // CHECK10-NEXT:    ret void
 //
 //
 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l535

 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK10:       omp.dispatch.cond:
 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK10-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK10-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP14]]
+// CHECK10-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP14]] to i32
+// CHECK10-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP13]], [[CONV7]]
 // CHECK10-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK10:       cond.true:
 // CHECK10-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK10-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP15]] to i32
 // CHECK10-NEXT:    br label [[COND_END:%.*]]
 // CHECK10:       cond.false:
 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK10-NEXT:    br label [[COND_END]]
 // CHECK10:       cond.end:
-// CHECK10-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK10-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK10-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
-// CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK10-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
+// CHECK10-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK10:       omp.dispatch.body:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK10:       omp.inner.for.cond:
 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !89
-// CHECK10-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
-// CHECK10-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK10-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
+// CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK10:       omp.inner.for.body:
 // CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !89
 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[TMP2]], align 8, !llvm.access.group !89
 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !89
 // CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i64 [[IDXPROM]]
 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !89
 // CHECK10-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[TMP3]], align 8, !llvm.access.group !89
 // CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !89
-// CHECK10-NEXT:    [[IDXPROM13:%.*]] = sext i32 [[TMP27]] to i64
-// CHECK10-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i64 [[IDXPROM13]]
-// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX14]], align 4, !llvm.access.group !89
-// CHECK10-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP25]], [[TMP28]]
+// CHECK10-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK10-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i64 [[IDXPROM12]]
+// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[ARRAYIDX13]], align 4, !llvm.access.group !89
+// CHECK10-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP25]], [[TMP28]]
 // CHECK10-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[TMP1]], align 8, !llvm.access.group !89
 // CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !89
-// CHECK10-NEXT:    [[IDXPROM16:%.*]] = sext i32 [[TMP30]] to i64
-// CHECK10-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[IDXPROM16]]
-// CHECK10-NEXT:    store i32 [[ADD15]], i32* [[ARRAYIDX17]], align 4, !llvm.access.group !89
+// CHECK10-NEXT:    [[IDXPROM15:%.*]] = sext i32 [[TMP30]] to i64
+// CHECK10-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[IDXPROM15]]
+// CHECK10-NEXT:    store i32 [[ADD14]], i32* [[ARRAYIDX16]], align 4, !llvm.access.group !89
 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK10:       omp.body.continue:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK10:       omp.inner.for.inc:
 // CHECK10-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
-// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP31]], 1
-// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
+// CHECK10-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP31]], 1
+// CHECK10-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !89
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP90:![0-9]+]]
 // CHECK10:       omp.inner.for.end:
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK10:       omp.dispatch.inc:
 // CHECK10-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK10-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
-// CHECK10-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP32]], [[TMP33]]
+// CHECK10-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK10-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
-// CHECK10-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
+// CHECK10-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK10:       omp.dispatch.end:
 // CHECK10-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP37]])
 // CHECK10-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK10-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 // CHECK10-NEXT:    br i1 [[TMP39]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK10:       .omp.final.then:
 // CHECK10-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK10-NEXT:    [[SUB21:%.*]] = sub nsw i32 [[TMP40]], 0
-// CHECK10-NEXT:    [[DIV22:%.*]] = sdiv i32 [[SUB21]], 1
-// CHECK10-NEXT:    [[MUL23:%.*]] = mul nsw i32 [[DIV22]], 1
-// CHECK10-NEXT:    [[ADD24:%.*]] = add nsw i32 0, [[MUL23]]
-// CHECK10-NEXT:    store i32 [[ADD24]], i32* [[I6]], align 4
+// CHECK10-NEXT:    [[SUB20:%.*]] = sub nsw i32 [[TMP40]], 0
+// CHECK10-NEXT:    [[DIV21:%.*]] = sdiv i32 [[SUB20]], 1
+// CHECK10-NEXT:    [[MUL22:%.*]] = mul nsw i32 [[DIV21]], 1
+// CHECK10-NEXT:    [[ADD23:%.*]] = add nsw i32 0, [[MUL22]]
+// CHECK10-NEXT:    store i32 [[ADD23]], i32* [[I6]], align 4
 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK10:       .omp.final.done:
 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK10:       omp.precond.end:
 // CHECK10-NEXT:    ret void
 //
 //
 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l82

 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK11:       omp.dispatch.cond:
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:

 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK11:       omp.dispatch.cond:
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:

 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK12-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK12:       omp.dispatch.cond:
 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK12-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK12-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK12-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK12:       cond.true:
 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK12-NEXT:    br label [[COND_END:%.*]]
 // CHECK12:       cond.false:
 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK12-NEXT:    br label [[COND_END]]
 // CHECK12:       cond.end:

 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP12]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP10]])
 // CHECK12-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK12:       omp.dispatch.cond:
 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK12-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP13]], [[TMP14]]
+// CHECK12-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK12-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK12:       cond.true:
 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK12-NEXT:    br label [[COND_END:%.*]]
 // CHECK12:       cond.false:
 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK12-NEXT:    br label [[COND_END]]
 // CHECK12:       cond.end:

clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_codegen.cpp

 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 32)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP10]]
+// CHECK1-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP10]] to i32
+// CHECK1-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP9]], [[CONV7]]
 // CHECK1-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP11]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP12]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP11]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
-// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
+// CHECK1-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
-// CHECK1-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
+// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 1, i32* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK1-NEXT:    store i32 [[TMP20]], i32* [[CONV1]], align 8
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP21]], 1
-// CHECK1-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP21]], 1
+// CHECK1-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
-// CHECK1-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
+// CHECK1-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
-// CHECK1-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
+// CHECK1-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]])
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 // CHECK1-NEXT:    br i1 [[TMP29]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]

 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 32)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP10]]
+// CHECK2-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP10]] to i32
+// CHECK2-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP9]], [[CONV7]]
 // CHECK2-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK2-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP11]] to i32
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP12]] to i64
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
-// CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP11]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK2-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK2-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
-// CHECK2-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
+// CHECK2-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
-// CHECK2-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
+// CHECK2-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK2-NEXT:    store i32 1, i32* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I6]], align 4
 // CHECK2-NEXT:    store i32 [[TMP20]], i32* [[CONV1]], align 8
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP21]], 1
-// CHECK2-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP21]], 1
+// CHECK2-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
-// CHECK2-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
+// CHECK2-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
-// CHECK2-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
+// CHECK2-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]])
 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 // CHECK2-NEXT:    br i1 [[TMP29]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]

 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 32)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP4:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:

 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 32)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK4-NEXT:    [[CMP4:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// CHECK4-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
 // CHECK4-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:

clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_generic_mode_codegen.cpp

 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV8:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP9:%.*]] = icmp ugt i64 [[CONV8]], [[TMP11]]
+// CHECK1-NEXT:    [[CONV8:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK1-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP10]], [[CONV8]]
 // CHECK1-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV10:%.*]] = trunc i64 [[TMP12]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV10:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE]] ], [ [[CONV10]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV11:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV11]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV10]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
-// CHECK1-NEXT:    br i1 [[CMP12]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP13:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
-// CHECK1-NEXT:    br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
+// CHECK1-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I7]], align 4
 // CHECK1-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I7]]) #[[ATTR4:[0-9]+]]
-// CHECK1-NEXT:    [[CALL14:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
-// CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CALL]], [[CALL14]]
-// CHECK1-NEXT:    [[CALL16:%.*]] = call i32 @_Z3fooPi(i32* [[CONV]]) #[[ATTR4]]
-// CHECK1-NEXT:    [[ADD17:%.*]] = add nsw i32 [[ADD15]], [[CALL16]]
-// CHECK1-NEXT:    store i32 [[ADD17]], i32* [[TMP0]], align 4
+// CHECK1-NEXT:    [[CALL13:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
+// CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[CALL]], [[CALL13]]
+// CHECK1-NEXT:    [[CALL15:%.*]] = call i32 @_Z3fooPi(i32* [[CONV]]) #[[ATTR4]]
+// CHECK1-NEXT:    [[ADD16:%.*]] = add nsw i32 [[ADD14]], [[CALL15]]
+// CHECK1-NEXT:    store i32 [[ADD16]], i32* [[TMP0]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP20]], 1
-// CHECK1-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
+// CHECK1-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
-// CHECK1-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
+// CHECK1-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
-// CHECK1-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
+// CHECK1-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void

 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK2-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP10]], [[TMP11]]
+// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
 // CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:

 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP10]], [[TMP11]]
+// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:

 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CONV8:%.*]] = sext i32 [[TMP10]] to i64
 // CHECK4-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK4-NEXT:    [[CMP9:%.*]] = icmp ugt i64 [[CONV8]], [[TMP11]]
+// CHECK4-NEXT:    [[CONV8:%.*]] = trunc i64 [[TMP11]] to i32
+// CHECK4-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP10]], [[CONV8]]
 // CHECK4-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
 // CHECK4-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK4-NEXT:    [[CONV10:%.*]] = trunc i64 [[TMP12]] to i32
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CONV10:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:
-// CHECK4-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE]] ], [ [[CONV10]], [[COND_FALSE]] ]
-// CHECK4-NEXT:    [[CONV11:%.*]] = trunc i64 [[COND]] to i32
-// CHECK4-NEXT:    store i32 [[CONV11]], i32* [[DOTOMP_UB]], align 4
+// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV10]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
-// CHECK4-NEXT:    br i1 [[CMP12]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK4-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// CHECK4-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK4:       omp.dispatch.body:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:
 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CMP13:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
-// CHECK4-NEXT:    br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK4-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
+// CHECK4-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK4:       omp.inner.for.body:
 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I7]], align 4
 // CHECK4-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooPi(i32* [[I7]]) #[[ATTR4:[0-9]+]]
-// CHECK4-NEXT:    [[CALL14:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
-// CHECK4-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CALL]], [[CALL14]]
-// CHECK4-NEXT:    [[CALL16:%.*]] = call i32 @_Z3fooPi(i32* [[CONV]]) #[[ATTR4]]
-// CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i32 [[ADD15]], [[CALL16]]
-// CHECK4-NEXT:    store i32 [[ADD17]], i32* [[TMP0]], align 4
+// CHECK4-NEXT:    [[CALL13:%.*]] = call i32 @_Z3fooPi(i32* [[TMP0]]) #[[ATTR4]]
+// CHECK4-NEXT:    [[ADD14:%.*]] = add nsw i32 [[CALL]], [[CALL13]]
+// CHECK4-NEXT:    [[CALL15:%.*]] = call i32 @_Z3fooPi(i32* [[CONV]]) #[[ATTR4]]
+// CHECK4-NEXT:    [[ADD16:%.*]] = add nsw i32 [[ADD14]], [[CALL15]]
+// CHECK4-NEXT:    store i32 [[ADD16]], i32* [[TMP0]], align 4
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK4:       omp.body.continue:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK4:       omp.inner.for.inc:
 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK4-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP20]], 1
-// CHECK4-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV]], align 4
+// CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
+// CHECK4-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK4:       omp.inner.for.end:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK4:       omp.dispatch.inc:
 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK4-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
-// CHECK4-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_LB]], align 4
+// CHECK4-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
+// CHECK4-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK4-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
-// CHECK4-NEXT:    store i32 [[ADD20]], i32* [[DOTOMP_UB]], align 4
+// CHECK4-NEXT:    [[ADD19:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
+// CHECK4-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK4:       omp.dispatch.end:
 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK4:       omp.precond.end:
 // CHECK4-NEXT:    ret void

 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK5:       omp.dispatch.cond:
 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK5-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP10]], [[TMP11]]
+// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
 // CHECK5-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK5:       cond.true:
 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK5-NEXT:    br label [[COND_END:%.*]]
 // CHECK5:       cond.false:
 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    br label [[COND_END]]
 // CHECK5:       cond.end:

 // CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK6:       omp.dispatch.cond:
 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK6-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP10]], [[TMP11]]
+// CHECK6-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
 // CHECK6-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK6:       cond.true:
 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK6-NEXT:    br label [[COND_END:%.*]]
 // CHECK6:       cond.false:
 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    br label [[COND_END]]
 // CHECK6:       cond.end:

clang/test/OpenMP/nvptx_target_teams_distribute_parallel_for_simd_codegen.cpp

 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK1-NEXT:    [[CMP7:%.*]] = icmp slt i32 [[TMP12]], [[ADD]]
 // CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !12
 // CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[N_CASTED]] to i32*
-// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[CONV8]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
-// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 8
+// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[CONV8]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 8, !llvm.access.group !12
 // CHECK1-NEXT:    [[CONV9:%.*]] = bitcast i64* [[L_CASTED]] to i32*
-// CHECK1-NEXT:    store i32 [[TMP20]], i32* [[CONV9]], align 4
-// CHECK1-NEXT:    [[TMP21:%.*]] = load i64, i64* [[L_CASTED]], align 8
+// CHECK1-NEXT:    store i32 [[TMP20]], i32* [[CONV9]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i64, i64* [[L_CASTED]], align 8, !llvm.access.group !12
 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP23:%.*]] = inttoptr i64 [[TMP15]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// CHECK1-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8, !llvm.access.group !12
 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK1-NEXT:    [[TMP25:%.*]] = inttoptr i64 [[TMP17]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8
+// CHECK1-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8, !llvm.access.group !12
 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP27:%.*]] = inttoptr i64 [[TMP19]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 8
+// CHECK1-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 8, !llvm.access.group !12
 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3
 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast [1000 x i32]* [[TMP0]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP29]], i8** [[TMP28]], align 8
+// CHECK1-NEXT:    store i8* [[TMP29]], i8** [[TMP28]], align 8, !llvm.access.group !12
 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 4
 // CHECK1-NEXT:    [[TMP31:%.*]] = inttoptr i64 [[TMP21]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP31]], i8** [[TMP30]], align 8
-// CHECK1-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
+// CHECK1-NEXT:    store i8* [[TMP31]], i8** [[TMP30]], align 8, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    [[TMP34:%.*]] = bitcast [5 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP33]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*, i64)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP34]], i64 5)
+// CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP33]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*, i64)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP34]], i64 5), !llvm.access.group !12
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP35]], [[TMP36]]
-// CHECK1-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP37]], [[TMP38]]
-// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP39]], [[TMP40]]
-// CHECK1-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
+// CHECK1-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    [[CMP13:%.*]] = icmp sgt i32 [[TMP41]], [[TMP42]]
 // CHECK1-NEXT:    br i1 [[CMP13]], label [[COND_TRUE14:%.*]], label [[COND_FALSE15:%.*]]
 // CHECK1:       cond.true14:
-// CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
+// CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    br label [[COND_END16:%.*]]
 // CHECK1:       cond.false15:
-// CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
 // CHECK1-NEXT:    br label [[COND_END16]]
 // CHECK1:       cond.end16:
 // CHECK1-NEXT:    [[COND17:%.*]] = phi i32 [ [[TMP43]], [[COND_TRUE14]] ], [ [[TMP44]], [[COND_FALSE15]] ]
-// CHECK1-NEXT:    store i32 [[COND17]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP45]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
+// CHECK1-NEXT:    store i32 [[COND17]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    store i32 [[TMP45]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP46:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP47:%.*]] = load i32, i32* [[TMP46]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP47]])
 // CHECK1-NEXT:    [[TMP48:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP49:%.*]] = icmp ne i32 [[TMP48]], 0

 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 32)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP8:%.*]] = icmp ugt i64 [[CONV7]], [[TMP10]]
+// CHECK1-NEXT:    [[CONV7:%.*]] = trunc i64 [[TMP10]] to i32
+// CHECK1-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP9]], [[CONV7]]
 // CHECK1-NEXT:    br i1 [[CMP8]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV9:%.*]] = trunc i64 [[TMP11]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV9:%.*]] = sext i32 [[TMP12]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP11]], [[COND_TRUE]] ], [ [[CONV9]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV10:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV10]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV9]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
-// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
+// CHECK1-NEXT:    br i1 [[CMP10]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP12:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
-// CHECK1-NEXT:    br i1 [[CMP12]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
+// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
+// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
+// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !16
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !16
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    store i32 1, i32* [[ARRAYIDX]], align 4
-// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I6]], align 4
-// CHECK1-NEXT:    store i32 [[TMP20]], i32* [[CONV1]], align 8
+// CHECK1-NEXT:    store i32 1, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !16
+// CHECK1-NEXT:    store i32 [[TMP20]], i32* [[CONV1]], align 8, !llvm.access.group !16
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP21]], 1
-// CHECK1-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
+// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP21]], 1
+// CHECK1-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
-// CHECK1-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
+// CHECK1-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
-// CHECK1-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
+// CHECK1-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]])
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 // CHECK1-NEXT:    br i1 [[TMP29]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[TMP30]], 0
-// CHECK1-NEXT:    [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
-// CHECK1-NEXT:    [[MUL18:%.*]] = mul nsw i32 [[DIV17]], 1
-// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 0, [[MUL18]]
-// CHECK1-NEXT:    store i32 [[ADD19]], i32* [[I6]], align 4
+// CHECK1-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP30]], 0
+// CHECK1-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
+// CHECK1-NEXT:    [[MUL17:%.*]] = mul nsw i32 [[DIV16]], 1
+// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 0, [[MUL17]]
+// CHECK1-NEXT:    store i32 [[ADD18]], i32* [[I6]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
 // CHECK1-NEXT:    br i1 [[TMP32]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 // CHECK1:       .omp.lastprivate.then:
 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[CONV1]], align 8
 // CHECK1-NEXT:    store i32 [[TMP33]], i32* [[CONV1]], align 8

 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[TMP12]], [[ADD]]
 // CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !19
 // CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
-// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
+// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8, !llvm.access.group !19
 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP21:%.*]] = inttoptr i64 [[TMP15]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 8
+// CHECK1-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 8, !llvm.access.group !19
 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK1-NEXT:    [[TMP23:%.*]] = inttoptr i64 [[TMP17]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// CHECK1-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8, !llvm.access.group !19
 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP25:%.*]] = inttoptr i64 [[TMP19]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8
+// CHECK1-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8, !llvm.access.group !19
 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3
 // CHECK1-NEXT:    [[TMP27:%.*]] = bitcast [1000 x i16]* [[TMP0]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 8
-// CHECK1-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
+// CHECK1-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 8, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    [[TMP30:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i16]*)* @__omp_outlined__3 to i8*), i8* null, i8** [[TMP30]], i64 4)
+// CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i16]*)* @__omp_outlined__3 to i8*), i8* null, i8** [[TMP30]], i64 4), !llvm.access.group !19
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
-// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
-// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP35]], [[TMP36]]
-// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP37]], [[TMP38]]
 // CHECK1-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
 // CHECK1:       cond.true11:
-// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    br label [[COND_END13:%.*]]
 // CHECK1:       cond.false12:
-// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
 // CHECK1-NEXT:    br label [[COND_END13]]
 // CHECK1:       cond.end13:
 // CHECK1-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP39]], [[COND_TRUE11]] ], [ [[TMP40]], [[COND_FALSE12]] ]
-// CHECK1-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP41]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
+// CHECK1-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    store i32 [[TMP41]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP42:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[TMP42]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP43]])
 // CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP45:%.*]] = icmp ne i32 [[TMP44]], 0

 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
 // CHECK1-NEXT:    [[CONV6:%.*]] = sext i32 [[TMP10]] to i64
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group !22
 // CHECK1-NEXT:    [[CMP7:%.*]] = icmp ule i64 [[CONV6]], [[TMP11]]
 // CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I5]], align 4
+// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4, !llvm.access.group !22
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I5]], align 4, !llvm.access.group !22
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i16], [1000 x i16]* [[TMP0]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i16, i16* [[ARRAYIDX]], align 2
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i16, i16* [[ARRAYIDX]], align 2, !llvm.access.group !22
 // CHECK1-NEXT:    [[CONV8:%.*]] = sext i16 [[TMP14]] to i32
 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[CONV8]], 1
 // CHECK1-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD9]] to i16
-// CHECK1-NEXT:    store i16 [[CONV10]], i16* [[ARRAYIDX]], align 2
+// CHECK1-NEXT:    store i16 [[CONV10]], i16* [[ARRAYIDX]], align 2, !llvm.access.group !22
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
 // CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
-// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
+// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP18]])
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0

 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 10
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP8]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 8
+// CHECK1-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 8, !llvm.access.group !25
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK1-NEXT:    [[TMP14:%.*]] = inttoptr i64 [[TMP10]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 8
+// CHECK1-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 8, !llvm.access.group !25
 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast [10 x i32]* [[TMP0]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8
+// CHECK1-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8, !llvm.access.group !25
 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP17]], i64 3)
+// CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP17]], i64 3), !llvm.access.group !25
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
-// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP24]], 9
 // CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
 // CHECK1:       cond.true5:
 // CHECK1-NEXT:    br label [[COND_END7:%.*]]
 // CHECK1:       cond.false6:
-// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK1-NEXT:    br label [[COND_END7]]
 // CHECK1:       cond.end7:
 // CHECK1-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP25]], [[COND_FALSE6]] ]
-// CHECK1-NEXT:    store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
+// CHECK1-NEXT:    store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
 // CHECK1-NEXT:    br i1 [[TMP28]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:

 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP6]] to i64
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group !28
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV2]], [[TMP7]]
 // 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, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4
+// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !28
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !28
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !28
 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
+// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !28
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
+// CHECK1-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:

 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 100
 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[CONV]], align 8
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !31
 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[F_CASTED]] to i32*
-// CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[F_CASTED]], align 8
+// CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[F_CASTED]], align 8, !llvm.access.group !31
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP14:%.*]] = inttoptr i64 [[TMP8]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 8
+// CHECK1-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 8, !llvm.access.group !31
 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
 // CHECK1-NEXT:    [[TMP16:%.*]] = inttoptr i64 [[TMP10]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8
+// CHECK1-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8, !llvm.access.group !31
 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast [10 x [10 x i32]]* [[TMP0]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 8
+// CHECK1-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 8, !llvm.access.group !31
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i64 0, i64 3
 // CHECK1-NEXT:    [[TMP20:%.*]] = inttoptr i64 [[TMP12]] to i8*
-// CHECK1-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
+// CHECK1-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8, !llvm.access.group !31
 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i64, i64, [10 x [10 x i32]]*, i64)* @__omp_outlined__7 to i8*), i8* null, i8** [[TMP21]], i64 4)
+// CHECK1-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i64, i64, [10 x [10 x i32]]*, i64)* @__omp_outlined__7 to i8*), i8* null, i8** [[TMP21]], i64 4), !llvm.access.group !31
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
-// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
-// CHECK1-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
-// CHECK1-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP28]], 99
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[COND_TRUE7:%.*]], label [[COND_FALSE8:%.*]]
 // CHECK1:       cond.true7:
 // CHECK1-NEXT:    br label [[COND_END9:%.*]]
 // CHECK1:       cond.false8:
-// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
 // CHECK1-NEXT:    br label [[COND_END9]]
 // CHECK1:       cond.end9:
 // CHECK1-NEXT:    [[COND10:%.*]] = phi i32 [ 99, [[COND_TRUE7]] ], [ [[TMP29]], [[COND_FALSE8]] ]
-// CHECK1-NEXT:    store i32 [[COND10]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP30]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
+// CHECK1-NEXT:    store i32 [[COND10]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    store i32 [[TMP30]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
 // CHECK1-NEXT:    br i1 [[TMP32]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:

 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
 // CHECK1-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP6]] to i64
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8, !llvm.access.group !34
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV4]], [[TMP7]]
 // 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, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 10
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !34
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
 // CHECK1-NEXT:    [[DIV5:%.*]] = sdiv i32 [[TMP10]], 10
 // CHECK1-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[DIV5]], 10
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL6]]
 // CHECK1-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL7]]
-// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[J]], align 4
-// CHECK1-NEXT:    store i32 10, i32* [[K]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[J]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[CONV]], align 8
+// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[J]], align 4, !llvm.access.group !34
+// CHECK1-NEXT:    store i32 10, i32* [[K]], align 4, !llvm.access.group !34
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !34
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !34
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[CONV]], align 8, !llvm.access.group !34
 // CHECK1-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[TMP12]], [[TMP13]]
 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP11]], [[MUL9]]
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[K]], align 4
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[K]], align 4, !llvm.access.group !34
 // CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD10]], [[TMP14]]
-// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !34
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], [10 x [10 x i32]]* [[TMP0]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[J]], align 4
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !34
 // CHECK1-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[ARRAYIDX]], i64 0, i64 [[IDXPROM12]]
-// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[ARRAYIDX13]], align 4
+// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[ARRAYIDX13]], align 4, !llvm.access.group !34
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
 // CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK1-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
+// CHECK1-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
 // CHECK1-NEXT:    br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:

 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK2-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP12]], [[ADD]]
 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
-// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
-// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
-// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[L_ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[L_CASTED]], align 4
-// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[L_CASTED]], align 4
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[L_ADDR]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[L_CASTED]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[L_CASTED]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP21:%.*]] = inttoptr i32 [[TMP14]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 4
+// CHECK2-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP23:%.*]] = inttoptr i32 [[TMP15]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 4
+// CHECK2-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP25:%.*]] = inttoptr i32 [[TMP17]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 4
+// CHECK2-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 3
 // CHECK2-NEXT:    [[TMP27:%.*]] = bitcast [1000 x i32]* [[TMP0]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 4
+// CHECK2-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 4
 // CHECK2-NEXT:    [[TMP29:%.*]] = inttoptr i32 [[TMP19]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP29]], i8** [[TMP28]], align 4
-// CHECK2-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4
+// CHECK2-NEXT:    store i8* [[TMP29]], i8** [[TMP28]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[TMP32:%.*]] = bitcast [5 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP31]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*, i32)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP32]], i32 5)
+// CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP31]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*, i32)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP32]], i32 5), !llvm.access.group !12
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
-// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP35]], [[TMP36]]
-// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP37]], [[TMP38]]
-// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP39]], [[TMP40]]
 // CHECK2-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
 // CHECK2:       cond.true11:
-// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    br label [[COND_END13:%.*]]
 // CHECK2:       cond.false12:
-// CHECK2-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK2-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
 // CHECK2-NEXT:    br label [[COND_END13]]
 // CHECK2:       cond.end13:
 // CHECK2-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP41]], [[COND_TRUE11]] ], [ [[TMP42]], [[COND_FALSE12]] ]
-// CHECK2-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    store i32 [[TMP43]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
+// CHECK2-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    store i32 [[TMP43]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    [[TMP44:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP45:%.*]] = load i32, i32* [[TMP44]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP45]])
 // CHECK2-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP47:%.*]] = icmp ne i32 [[TMP46]], 0

 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 32)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK2-NEXT:    [[CMP4:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
 // CHECK2-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
 // CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
-// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
-// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
+// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !16
+// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP19]]
-// CHECK2-NEXT:    store i32 1, i32* [[ARRAYIDX]], align 4
-// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4
-// CHECK2-NEXT:    store i32 [[TMP20]], i32* [[L_ADDR]], align 4
+// CHECK2-NEXT:    store i32 1, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
+// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
+// CHECK2-NEXT:    store i32 [[TMP20]], i32* [[L_ADDR]], align 4, !llvm.access.group !16
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], 1
-// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
+// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
 // CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4

 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK2-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[TMP12]], [[ADD]]
 // CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
-// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
-// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP19:%.*]] = inttoptr i32 [[TMP14]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP19]], i8** [[TMP18]], align 4
+// CHECK2-NEXT:    store i8* [[TMP19]], i8** [[TMP18]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP21:%.*]] = inttoptr i32 [[TMP15]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 4
+// CHECK2-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP23:%.*]] = inttoptr i32 [[TMP17]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 4
+// CHECK2-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 3
 // CHECK2-NEXT:    [[TMP25:%.*]] = bitcast [1000 x i16]* [[TMP0]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 4
-// CHECK2-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
+// CHECK2-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[TMP28:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP27]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i16]*)* @__omp_outlined__3 to i8*), i8* null, i8** [[TMP28]], i32 4)
+// CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP27]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i16]*)* @__omp_outlined__3 to i8*), i8* null, i8** [[TMP28]], i32 4), !llvm.access.group !19
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP29]], [[TMP30]]
-// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
-// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
-// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP35]], [[TMP36]]
 // CHECK2-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
 // CHECK2:       cond.true10:
-// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    br label [[COND_END12:%.*]]
 // CHECK2:       cond.false11:
-// CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
 // CHECK2-NEXT:    br label [[COND_END12]]
 // CHECK2:       cond.end12:
 // CHECK2-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP37]], [[COND_TRUE10]] ], [ [[TMP38]], [[COND_FALSE11]] ]
-// CHECK2-NEXT:    store i32 [[COND13]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    store i32 [[TMP39]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
+// CHECK2-NEXT:    store i32 [[COND13]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    store i32 [[TMP39]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP41]])
 // CHECK2-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0

 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group !22
 // CHECK2-NEXT:    [[CMP4:%.*]] = icmp ule 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, i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
-// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I3]], align 4
+// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !22
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !22
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i16], [1000 x i16]* [[TMP0]], i32 0, i32 [[TMP13]]
-// CHECK2-NEXT:    [[TMP14:%.*]] = load i16, i16* [[ARRAYIDX]], align 2
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i16, i16* [[ARRAYIDX]], align 2, !llvm.access.group !22
 // CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP14]] to i32
 // CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV]], 1
 // CHECK2-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
-// CHECK2-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX]], align 2
+// CHECK2-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX]], align 2, !llvm.access.group !22
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
-// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
+// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP18]])
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0

 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 10
 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
-// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP10:%.*]] = inttoptr i32 [[TMP7]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
+// CHECK2-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP8]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4
+// CHECK2-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP14:%.*]] = bitcast [10 x i32]* [[TMP0]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4
+// CHECK2-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP15]], i32 3)
+// CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP15]], i32 3), !llvm.access.group !25
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
-// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP22]], 9
 // CHECK2-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
 // CHECK2:       cond.true5:
 // CHECK2-NEXT:    br label [[COND_END7:%.*]]
 // CHECK2:       cond.false6:
-// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK2-NEXT:    br label [[COND_END7]]
 // CHECK2:       cond.end7:
 // CHECK2-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP23]], [[COND_FALSE6]] ]
-// CHECK2-NEXT:    store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
+// CHECK2-NEXT:    store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
 // CHECK2-NEXT:    br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK2:       .omp.final.then:

 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group !28
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp ule 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, i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
-// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4
+// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !28
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !28
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
-// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !28
 // CHECK2-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK2-NEXT:    store i32 [[ADD1]], i32* [[ARRAYIDX]], align 4
+// CHECK2-NEXT:    store i32 [[ADD1]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !28
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
+// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK2:       .omp.final.then:

 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 100
 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
-// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[F_ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[TMP9]], i32* [[F_CASTED]], align 4
-// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[F_CASTED]], align 4
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[F_ADDR]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    store i32 [[TMP9]], i32* [[F_CASTED]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[F_CASTED]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP7]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4
+// CHECK2-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK2-NEXT:    [[TMP14:%.*]] = inttoptr i32 [[TMP8]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4
+// CHECK2-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK2-NEXT:    [[TMP16:%.*]] = bitcast [10 x [10 x i32]]* [[TMP0]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 4
+// CHECK2-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 3
 // CHECK2-NEXT:    [[TMP18:%.*]] = inttoptr i32 [[TMP10]] to i8*
-// CHECK2-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 4
+// CHECK2-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[TMP19:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, [10 x [10 x i32]]*, i32)* @__omp_outlined__7 to i8*), i8* null, i8** [[TMP19]], i32 4)
+// CHECK2-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, [10 x [10 x i32]]*, i32)* @__omp_outlined__7 to i8*), i8* null, i8** [[TMP19]], i32 4), !llvm.access.group !31
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
-// CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
-// CHECK2-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK2-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP26]], 99
 // CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
 // CHECK2:       cond.true6:
 // CHECK2-NEXT:    br label [[COND_END8:%.*]]
 // CHECK2:       cond.false7:
-// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
 // CHECK2-NEXT:    br label [[COND_END8]]
 // CHECK2:       cond.end8:
 // CHECK2-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP27]], [[COND_FALSE7]] ]
-// CHECK2-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
+// CHECK2-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
 // CHECK2-NEXT:    br i1 [[TMP30]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK2:       .omp.final.then:

 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp ule 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, i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 10
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
-// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    [[DIV2:%.*]] = sdiv i32 [[TMP10]], 10
 // CHECK2-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 10
 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL3]]
 // CHECK2-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 0, [[MUL4]]
-// CHECK2-NEXT:    store i32 [[ADD5]], i32* [[J]], align 4
-// CHECK2-NEXT:    store i32 10, i32* [[K]], align 4
-// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
-// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[J]], align 4
-// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[F_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[ADD5]], i32* [[J]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    store i32 10, i32* [[K]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[F_ADDR]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[TMP12]], [[TMP13]]
 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], [[MUL6]]
-// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[K]], align 4
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[K]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[ADD7]], [[TMP14]]
-// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], [10 x [10 x i32]]* [[TMP0]], i32 0, i32 [[TMP15]]
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[J]], align 4
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[ARRAYIDX]], i32 0, i32 [[TMP16]]
-// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX9]], align 4
+// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX9]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
 // CHECK2-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK2-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
+// CHECK2-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
 // CHECK2-NEXT:    br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK2:       .omp.final.then:

 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP12]], [[ADD]]
 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[L_ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP18]], i32* [[L_CASTED]], align 4
-// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[L_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[L_ADDR]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    store i32 [[TMP18]], i32* [[L_CASTED]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[L_CASTED]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP21:%.*]] = inttoptr i32 [[TMP14]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 4
+// CHECK3-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP23:%.*]] = inttoptr i32 [[TMP15]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 4
+// CHECK3-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP25:%.*]] = inttoptr i32 [[TMP17]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 4
+// CHECK3-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 3
 // CHECK3-NEXT:    [[TMP27:%.*]] = bitcast [1000 x i32]* [[TMP0]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 4
+// CHECK3-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 4
 // CHECK3-NEXT:    [[TMP29:%.*]] = inttoptr i32 [[TMP19]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP29]], i8** [[TMP28]], align 4
-// CHECK3-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4
+// CHECK3-NEXT:    store i8* [[TMP29]], i8** [[TMP28]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[TMP32:%.*]] = bitcast [5 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP31]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*, i32)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP32]], i32 5)
+// CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP31]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*, i32)* @__omp_outlined__1 to i8*), i8* null, i8** [[TMP32]], i32 5), !llvm.access.group !12
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
-// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP35]], [[TMP36]]
-// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP37]], [[TMP38]]
-// CHECK3-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK3-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP39]], [[TMP40]]
 // CHECK3-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
 // CHECK3:       cond.true11:
-// CHECK3-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK3-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    br label [[COND_END13:%.*]]
 // CHECK3:       cond.false12:
-// CHECK3-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
 // CHECK3-NEXT:    br label [[COND_END13]]
 // CHECK3:       cond.end13:
 // CHECK3-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP41]], [[COND_TRUE11]] ], [ [[TMP42]], [[COND_FALSE12]] ]
-// CHECK3-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP43]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
+// CHECK3-NEXT:    store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    store i32 [[TMP43]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP44:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP45:%.*]] = load i32, i32* [[TMP44]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP45]])
 // CHECK3-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP47:%.*]] = icmp ne i32 [[TMP46]], 0

 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 32)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP4:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK3:       omp.dispatch.body:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
-// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
+// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !16
+// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP19]]
-// CHECK3-NEXT:    store i32 1, i32* [[ARRAYIDX]], align 4
-// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4
-// CHECK3-NEXT:    store i32 [[TMP20]], i32* [[L_ADDR]], align 4
+// CHECK3-NEXT:    store i32 1, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
+// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
+// CHECK3-NEXT:    store i32 [[TMP20]], i32* [[L_ADDR]], align 4, !llvm.access.group !16
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], 1
-// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
+// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK3:       omp.dispatch.inc:
 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
 // CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4

 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[TMP12]], [[ADD]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP19:%.*]] = inttoptr i32 [[TMP14]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP19]], i8** [[TMP18]], align 4
+// CHECK3-NEXT:    store i8* [[TMP19]], i8** [[TMP18]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP21:%.*]] = inttoptr i32 [[TMP15]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 4
+// CHECK3-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP23:%.*]] = inttoptr i32 [[TMP17]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 4
+// CHECK3-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 3
 // CHECK3-NEXT:    [[TMP25:%.*]] = bitcast [1000 x i16]* [[TMP0]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 4
-// CHECK3-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
+// CHECK3-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[TMP28:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP27]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i16]*)* @__omp_outlined__3 to i8*), i8* null, i8** [[TMP28]], i32 4)
+// CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP27]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i16]*)* @__omp_outlined__3 to i8*), i8* null, i8** [[TMP28]], i32 4), !llvm.access.group !19
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP29]], [[TMP30]]
-// CHECK3-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
-// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
-// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP35]], [[TMP36]]
 // CHECK3-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
 // CHECK3:       cond.true10:
-// CHECK3-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    br label [[COND_END12:%.*]]
 // CHECK3:       cond.false11:
-// CHECK3-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
 // CHECK3-NEXT:    br label [[COND_END12]]
 // CHECK3:       cond.end12:
 // CHECK3-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP37]], [[COND_TRUE10]] ], [ [[TMP38]], [[COND_FALSE11]] ]
-// CHECK3-NEXT:    store i32 [[COND13]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP39]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
+// CHECK3-NEXT:    store i32 [[COND13]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    store i32 [[TMP39]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP41]])
 // CHECK3-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0

 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group !22
 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp ule 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, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I3]], align 4
+// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !22
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !22
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i16], [1000 x i16]* [[TMP0]], i32 0, i32 [[TMP13]]
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i16, i16* [[ARRAYIDX]], align 2
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i16, i16* [[ARRAYIDX]], align 2, !llvm.access.group !22
 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP14]] to i32
 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV]], 1
 // CHECK3-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
-// CHECK3-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX]], align 2
+// CHECK3-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX]], align 2, !llvm.access.group !22
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
-// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
+// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP18]])
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0

 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 10
 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP10:%.*]] = inttoptr i32 [[TMP7]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4
+// CHECK3-NEXT:    store i8* [[TMP10]], i8** [[TMP9]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP8]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4
+// CHECK3-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast [10 x i32]* [[TMP0]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4
+// CHECK3-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[TMP15:%.*]] = bitcast [3 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP15]], i32 3)
+// CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @__omp_outlined__5 to i8*), i8* null, i8** [[TMP15]], i32 3), !llvm.access.group !25
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
-// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP22]], 9
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
 // CHECK3:       cond.true5:
 // CHECK3-NEXT:    br label [[COND_END7:%.*]]
 // CHECK3:       cond.false6:
-// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
 // CHECK3-NEXT:    br label [[COND_END7]]
 // CHECK3:       cond.end7:
 // CHECK3-NEXT:    [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP23]], [[COND_FALSE6]] ]
-// CHECK3-NEXT:    store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
+// CHECK3-NEXT:    store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
 // CHECK3-NEXT:    br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:

 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
+// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group !28
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ule 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, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4
+// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !28
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !28
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !28
 // CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK3-NEXT:    store i32 [[ADD1]], i32* [[ARRAYIDX]], align 4
+// CHECK3-NEXT:    store i32 [[ADD1]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !28
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !28
 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
+// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:

 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP6]], 100
 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[F_ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[F_CASTED]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[F_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[F_ADDR]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[F_CASTED]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[F_CASTED]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP12:%.*]] = inttoptr i32 [[TMP7]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4
+// CHECK3-NEXT:    store i8* [[TMP12]], i8** [[TMP11]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
 // CHECK3-NEXT:    [[TMP14:%.*]] = inttoptr i32 [[TMP8]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4
+// CHECK3-NEXT:    store i8* [[TMP14]], i8** [[TMP13]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast [10 x [10 x i32]]* [[TMP0]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 4
+// CHECK3-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[CAPTURED_VARS_ADDRS]], i32 0, i32 3
 // CHECK3-NEXT:    [[TMP18:%.*]] = inttoptr i32 [[TMP10]] to i8*
-// CHECK3-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 4
+// CHECK3-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast [4 x i8*]* [[CAPTURED_VARS_ADDRS]] to i8**
-// CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, [10 x [10 x i32]]*, i32)* @__omp_outlined__7 to i8*), i8* null, i8** [[TMP19]], i32 4)
+// CHECK3-NEXT:    call void @__kmpc_parallel_51(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*, i32, i32, [10 x [10 x i32]]*, i32)* @__omp_outlined__7 to i8*), i8* null, i8** [[TMP19]], i32 4), !llvm.access.group !31
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
-// CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
-// CHECK3-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP26]], 99
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
 // CHECK3:       cond.true6:
 // CHECK3-NEXT:    br label [[COND_END8:%.*]]
 // CHECK3:       cond.false7:
-// CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
 // CHECK3-NEXT:    br label [[COND_END8]]
 // CHECK3:       cond.end8:
 // CHECK3-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP27]], [[COND_FALSE7]] ]
-// CHECK3-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
+// CHECK3-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
 // CHECK3-NEXT:    br i1 [[TMP30]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:

 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ule 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, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 10
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    [[DIV2:%.*]] = sdiv i32 [[TMP10]], 10
 // CHECK3-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 10
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL3]]
 // CHECK3-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 0, [[MUL4]]
-// CHECK3-NEXT:    store i32 [[ADD5]], i32* [[J]], align 4
-// CHECK3-NEXT:    store i32 10, i32* [[K]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[J]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[F_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[ADD5]], i32* [[J]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    store i32 10, i32* [[K]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[F_ADDR]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[TMP12]], [[TMP13]]
 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], [[MUL6]]
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[K]], align 4
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[K]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[ADD7]], [[TMP14]]
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], [10 x [10 x i32]]* [[TMP0]], i32 0, i32 [[TMP15]]
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[J]], align 4
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[ARRAYIDX]], i32 0, i32 [[TMP16]]
-// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX9]], align 4
+// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX9]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !34
 // CHECK3-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK3-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
+// CHECK3-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
 // CHECK3-NEXT:    br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    store i32 10, i32* [[I]], align 4
 // CHECK3-NEXT:    store i32 10, i32* [[J]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    ret void
 //

clang/test/OpenMP/target_teams_distribute_parallel_for_ast_print.cpp

-// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=50 -ast-print %s -Wno-openmp-mapping | FileCheck %s
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=50 -ast-print %s -Wno-openmp-mapping -Wsign-conversion | FileCheck %s
 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -x c++ -std=c++11 -emit-pch -o %t %s -Wno-openmp-mapping
 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -std=c++11 -include-pch %t -fsyntax-only -verify %s -ast-print -Wno-openmp-mapping | FileCheck %s
 
-// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=50 -ast-print %s -Wno-openmp-mapping | FileCheck %s
+// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=50 -ast-print %s -Wno-openmp-mapping -Wsign-conversion | FileCheck %s
 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=50 -x c++ -std=c++11 -emit-pch -o %t %s -Wno-openmp-mapping
 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=50 -std=c++11 -include-pch %t -fsyntax-only -verify %s -ast-print -Wno-openmp-mapping | FileCheck %s
 // expected-no-diagnostics
 
 #ifndef HEADER
 #define HEADER
 
 typedef void **omp_allocator_handle_t;

clang/test/OpenMP/target_teams_distribute_parallel_for_schedule_codegen.cpp

 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l49
 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {

 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
-// CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK2-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK2-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l49
 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {

 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:

 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK4-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:

 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK5:       omp.dispatch.cond:
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK5-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK5-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK5:       cond.true:
 // CHECK5-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK5-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK5-NEXT:    br label [[COND_END:%.*]]
 // CHECK5:       cond.false:
 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK5-NEXT:    br label [[COND_END]]
 // CHECK5:       cond.end:
-// CHECK5-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK5-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK5-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK5-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK5-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK5:       omp.dispatch.body:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK5:       omp.inner.for.cond:
 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK5-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK5:       omp.body.continue:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK5:       omp.inner.for.inc:
 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK5-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK5-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
+// CHECK5-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK5-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK5:       omp.inner.for.end:
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK5:       omp.dispatch.inc:
 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK5-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK5-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK5-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK5-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK5-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK5-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK5:       omp.dispatch.end:
 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l49
 // CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {

 // CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK6:       omp.dispatch.cond:
 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK6-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK6-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK6-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK6:       cond.true:
 // CHECK6-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK6-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK6-NEXT:    br label [[COND_END:%.*]]
 // CHECK6:       cond.false:
 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK6-NEXT:    br label [[COND_END]]
 // CHECK6:       cond.end:
-// CHECK6-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK6-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK6-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK6-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK6-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK6-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK6:       omp.dispatch.body:
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK6:       omp.inner.for.cond:
 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK6-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK6-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK6-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK6:       omp.inner.for.body:
 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
 // CHECK6-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK6:       omp.body.continue:
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK6:       omp.inner.for.inc:
 // CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK6-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK6-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
+// CHECK6-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK6-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK6:       omp.inner.for.end:
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK6:       omp.dispatch.inc:
 // CHECK6-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK6-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK6-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK6-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK6-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK6-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK6-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK6-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK6-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK6:       omp.dispatch.end:
 // CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK6-NEXT:    ret void
 //
 //
 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l49
 // CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {

 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK7:       omp.dispatch.cond:
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK7-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK7:       cond.true:
 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK7-NEXT:    br label [[COND_END:%.*]]
 // CHECK7:       cond.false:
 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    br label [[COND_END]]
 // CHECK7:       cond.end:

 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK8-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK8:       omp.dispatch.cond:
 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK8-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK8:       cond.true:
 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK8-NEXT:    br label [[COND_END:%.*]]
 // CHECK8:       cond.false:
 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    br label [[COND_END]]
 // CHECK8:       cond.end:

 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK13:       omp.dispatch.cond:
 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK13-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK13-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK13:       cond.true:
 // CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK13-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK13-NEXT:    br label [[COND_END:%.*]]
 // CHECK13:       cond.false:
 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK13-NEXT:    br label [[COND_END]]
 // CHECK13:       cond.end:
-// CHECK13-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK13-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK13-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK13:       omp.dispatch.body:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK13:       omp.inner.for.cond:
 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK13-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK13:       omp.inner.for.body:
 // CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
 // CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK13:       omp.body.continue:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK13:       omp.inner.for.inc:
 // CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK13:       omp.inner.for.end:
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK13:       omp.dispatch.inc:
 // CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK13:       omp.dispatch.end:
 // CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK13-NEXT:    ret void
 //
 //
 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l124
 // CHECK13-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {

 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK14-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK14:       omp.dispatch.cond:
 // CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK14-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK14:       cond.true:
 // CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK14-NEXT:    br label [[COND_END:%.*]]
 // CHECK14:       cond.false:
 // CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK14-NEXT:    br label [[COND_END]]
 // CHECK14:       cond.end:
-// CHECK14-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK14-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK14-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK14:       omp.dispatch.body:
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK14:       omp.inner.for.cond:
 // CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK14-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK14:       omp.inner.for.body:
 // CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
 // CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK14:       omp.body.continue:
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK14:       omp.inner.for.inc:
 // CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK14:       omp.inner.for.end:
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK14:       omp.dispatch.inc:
 // CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK14:       omp.dispatch.end:
 // CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK14-NEXT:    ret void
 //
 //
 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l124
 // CHECK14-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {

 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK15:       omp.dispatch.cond:
 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK15-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK15:       cond.true:
 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK15-NEXT:    br label [[COND_END:%.*]]
 // CHECK15:       cond.false:
 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    br label [[COND_END]]
 // CHECK15:       cond.end:

 // CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK16-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK16:       omp.dispatch.cond:
 // CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK16-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK16:       cond.true:
 // CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK16-NEXT:    br label [[COND_END:%.*]]
 // CHECK16:       cond.false:
 // CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    br label [[COND_END]]
 // CHECK16:       cond.end:

 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK17:       omp.dispatch.cond:
 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK17-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK17-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK17-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK17:       cond.true:
 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK17-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK17-NEXT:    br label [[COND_END:%.*]]
 // CHECK17:       cond.false:
 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK17-NEXT:    br label [[COND_END]]
 // CHECK17:       cond.end:
-// CHECK17-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK17-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK17-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK17:       omp.dispatch.body:
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK17:       omp.inner.for.cond:
 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK17-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK17:       omp.inner.for.body:
 // CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
 // CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK17:       omp.body.continue:
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK17:       omp.inner.for.inc:
 // CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK17:       omp.inner.for.end:
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK17:       omp.dispatch.inc:
 // CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK17:       omp.dispatch.end:
 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK17-NEXT:    ret void
 //
 //
 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l124
 // CHECK17-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {

 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK18:       omp.dispatch.cond:
 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK18-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK18-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK18-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK18:       cond.true:
 // CHECK18-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK18-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK18-NEXT:    br label [[COND_END:%.*]]
 // CHECK18:       cond.false:
 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK18-NEXT:    br label [[COND_END]]
 // CHECK18:       cond.end:
-// CHECK18-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK18-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK18-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK18:       omp.dispatch.body:
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK18:       omp.inner.for.cond:
 // CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK18-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK18:       omp.inner.for.body:
 // CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
 // CHECK18-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK18:       omp.body.continue:
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK18:       omp.inner.for.inc:
 // CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK18-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK18:       omp.inner.for.end:
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK18:       omp.dispatch.inc:
 // CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK18-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK18-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK18:       omp.dispatch.end:
 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK18-NEXT:    ret void
 //
 //
 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l124
 // CHECK18-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {

 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK19:       omp.dispatch.cond:
 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK19-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK19:       cond.true:
 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK19-NEXT:    br label [[COND_END:%.*]]
 // CHECK19:       cond.false:
 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK19-NEXT:    br label [[COND_END]]
 // CHECK19:       cond.end:

 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK20:       omp.dispatch.cond:
 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK20-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK20:       cond.true:
 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK20-NEXT:    br label [[COND_END:%.*]]
 // CHECK20:       cond.false:
 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK20-NEXT:    br label [[COND_END]]
 // CHECK20:       cond.end:

clang/test/OpenMP/target_teams_distribute_parallel_for_simd_schedule_codegen.cpp

 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK1-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    store i32 123, i32* [[I]], align 4

 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
-// CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK2-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK2-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
-// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
-// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
+// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK2-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK2:       .omp.final.then:
 // CHECK2-NEXT:    store i32 123, i32* [[I]], align 4

 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:

 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK4-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:

 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK5:       omp.dispatch.cond:
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK5-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK5-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK5:       cond.true:
 // CHECK5-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK5-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK5-NEXT:    br label [[COND_END:%.*]]
 // CHECK5:       cond.false:
 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK5-NEXT:    br label [[COND_END]]
 // CHECK5:       cond.end:
-// CHECK5-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK5-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK5-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK5-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK5-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK5:       omp.dispatch.body:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK5:       omp.inner.for.cond:
 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
-// CHECK5-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK5-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
 // CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK5:       omp.body.continue:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK5:       omp.inner.for.inc:
 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
-// CHECK5-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK5-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
+// CHECK5-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK5-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
 // CHECK5:       omp.inner.for.end:
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK5:       omp.dispatch.inc:
 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK5-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK5-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK5-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK5-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK5-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK5-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK5:       omp.dispatch.end:
 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK5-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK5-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK5-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK5:       .omp.final.then:
 // CHECK5-NEXT:    store i32 123, i32* [[I]], align 4

 // CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK6:       omp.dispatch.cond:
 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK6-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK6-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK6-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK6:       cond.true:
 // CHECK6-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK6-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK6-NEXT:    br label [[COND_END:%.*]]
 // CHECK6:       cond.false:
 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK6-NEXT:    br label [[COND_END]]
 // CHECK6:       cond.end:
-// CHECK6-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK6-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK6-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK6-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK6-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK6-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK6:       omp.dispatch.body:
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK6:       omp.inner.for.cond:
 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
-// CHECK6-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK6-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK6-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK6-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK6:       omp.inner.for.body:
 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
 // CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
 // CHECK6-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK6:       omp.body.continue:
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK6:       omp.inner.for.inc:
 // CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
-// CHECK6-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK6-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
+// CHECK6-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK6-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
 // CHECK6:       omp.inner.for.end:
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK6:       omp.dispatch.inc:
 // CHECK6-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK6-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK6-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK6-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK6-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK6-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK6-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK6-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK6-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK6:       omp.dispatch.end:
 // CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK6-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK6-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK6-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK6:       .omp.final.then:
 // CHECK6-NEXT:    store i32 123, i32* [[I]], align 4

 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK7:       omp.dispatch.cond:
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK7-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK7:       cond.true:
 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK7-NEXT:    br label [[COND_END:%.*]]
 // CHECK7:       cond.false:
 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    br label [[COND_END]]
 // CHECK7:       cond.end:

 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK8-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK8:       omp.dispatch.cond:
 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK8-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK8:       cond.true:
 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK8-NEXT:    br label [[COND_END:%.*]]
 // CHECK8:       cond.false:
 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    br label [[COND_END]]
 // CHECK8:       cond.end:

 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK13:       omp.dispatch.cond:
 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK13-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK13-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK13:       cond.true:
 // CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK13-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK13-NEXT:    br label [[COND_END:%.*]]
 // CHECK13:       cond.false:
 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK13-NEXT:    br label [[COND_END]]
 // CHECK13:       cond.end:
-// CHECK13-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK13-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK13-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK13:       omp.dispatch.body:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK13:       omp.inner.for.cond:
 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
-// CHECK13-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK13-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK13:       omp.inner.for.body:
 // CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK13:       omp.body.continue:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK13:       omp.inner.for.inc:
 // CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
-// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
+// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
 // CHECK13:       omp.inner.for.end:
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK13:       omp.dispatch.inc:
 // CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK13:       omp.dispatch.end:
 // CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK13-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK13-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK13:       .omp.final.then:
 // CHECK13-NEXT:    store i32 10, i32* [[I]], align 4

 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK14-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK14:       omp.dispatch.cond:
 // CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK14-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK14:       cond.true:
 // CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK14-NEXT:    br label [[COND_END:%.*]]
 // CHECK14:       cond.false:
 // CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK14-NEXT:    br label [[COND_END]]
 // CHECK14:       cond.end:
-// CHECK14-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK14-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK14-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK14:       omp.dispatch.body:
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK14:       omp.inner.for.cond:
 // CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
-// CHECK14-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK14-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK14:       omp.inner.for.body:
 // CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK14:       omp.body.continue:
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK14:       omp.inner.for.inc:
 // CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
-// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
+// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
 // CHECK14:       omp.inner.for.end:
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK14:       omp.dispatch.inc:
 // CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK14:       omp.dispatch.end:
 // CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK14-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK14-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK14:       .omp.final.then:
 // CHECK14-NEXT:    store i32 10, i32* [[I]], align 4

 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK15:       omp.dispatch.cond:
 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK15-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK15:       cond.true:
 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK15-NEXT:    br label [[COND_END:%.*]]
 // CHECK15:       cond.false:
 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    br label [[COND_END]]
 // CHECK15:       cond.end:

 // CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK16-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK16:       omp.dispatch.cond:
 // CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK16-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK16:       cond.true:
 // CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK16-NEXT:    br label [[COND_END:%.*]]
 // CHECK16:       cond.false:
 // CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    br label [[COND_END]]
 // CHECK16:       cond.end:

 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK17:       omp.dispatch.cond:
 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK17-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK17-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK17-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK17:       cond.true:
 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK17-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK17-NEXT:    br label [[COND_END:%.*]]
 // CHECK17:       cond.false:
 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK17-NEXT:    br label [[COND_END]]
 // CHECK17:       cond.end:
-// CHECK17-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK17-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK17-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK17:       omp.dispatch.body:
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK17:       omp.inner.for.cond:
 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
-// CHECK17-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK17-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK17:       omp.inner.for.body:
 // CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
 // CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
 // CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK17:       omp.body.continue:
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK17:       omp.inner.for.inc:
 // CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
-// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
+// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
 // CHECK17:       omp.inner.for.end:
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK17:       omp.dispatch.inc:
 // CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK17:       omp.dispatch.end:
 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK17-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK17:       .omp.final.then:
 // CHECK17-NEXT:    store i32 10, i32* [[I]], align 4

 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK18:       omp.dispatch.cond:
 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK18-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK18-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK18-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK18:       cond.true:
 // CHECK18-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK18-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK18-NEXT:    br label [[COND_END:%.*]]
 // CHECK18:       cond.false:
 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK18-NEXT:    br label [[COND_END]]
 // CHECK18:       cond.end:
-// CHECK18-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK18-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK18-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK18:       omp.dispatch.body:
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK18:       omp.inner.for.cond:
 // CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
-// CHECK18-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK18-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK18:       omp.inner.for.body:
 // CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
 // CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
 // CHECK18-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK18:       omp.body.continue:
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK18:       omp.inner.for.inc:
 // CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
-// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
+// CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK18-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
 // CHECK18:       omp.inner.for.end:
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK18:       omp.dispatch.inc:
 // CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK18-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK18-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK18:       omp.dispatch.end:
 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK18-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK18-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK18:       .omp.final.then:
 // CHECK18-NEXT:    store i32 10, i32* [[I]], align 4

 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK19:       omp.dispatch.cond:
 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK19-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK19:       cond.true:
 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK19-NEXT:    br label [[COND_END:%.*]]
 // CHECK19:       cond.false:
 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK19-NEXT:    br label [[COND_END]]
 // CHECK19:       cond.end:

 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK20:       omp.dispatch.cond:
 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK20-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK20:       cond.true:
 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK20-NEXT:    br label [[COND_END:%.*]]
 // CHECK20:       cond.false:
 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK20-NEXT:    br label [[COND_END]]
 // CHECK20:       cond.end:

clang/test/OpenMP/teams_distribute_parallel_for_schedule_codegen.cpp

 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {

 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
-// CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK2-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK2-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {

 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:

 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK4-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:

 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK5:       omp.dispatch.cond:
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK5-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK5-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK5:       cond.true:
 // CHECK5-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK5-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK5-NEXT:    br label [[COND_END:%.*]]
 // CHECK5:       cond.false:
 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK5-NEXT:    br label [[COND_END]]
 // CHECK5:       cond.end:
-// CHECK5-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK5-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK5-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK5-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK5-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK5:       omp.dispatch.body:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK5:       omp.inner.for.cond:
 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK5-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK5:       omp.body.continue:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK5:       omp.inner.for.inc:
 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK5-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK5-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
+// CHECK5-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK5-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK5:       omp.inner.for.end:
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK5:       omp.dispatch.inc:
 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK5-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK5-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK5-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK5-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK5-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK5-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK5:       omp.dispatch.end:
 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
 // CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {

 // CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK6:       omp.dispatch.cond:
 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK6-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK6-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK6-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK6:       cond.true:
 // CHECK6-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK6-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK6-NEXT:    br label [[COND_END:%.*]]
 // CHECK6:       cond.false:
 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK6-NEXT:    br label [[COND_END]]
 // CHECK6:       cond.end:
-// CHECK6-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK6-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK6-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK6-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK6-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK6-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK6:       omp.dispatch.body:
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK6:       omp.inner.for.cond:
 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK6-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK6-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK6-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK6-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK6:       omp.inner.for.body:
 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK6-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
 // CHECK6-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK6:       omp.body.continue:
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK6:       omp.inner.for.inc:
 // CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK6-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK6-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
+// CHECK6-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK6-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK6:       omp.inner.for.end:
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK6:       omp.dispatch.inc:
 // CHECK6-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK6-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK6-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK6-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK6-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK6-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK6-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK6-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK6-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK6-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK6:       omp.dispatch.end:
 // CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK6-NEXT:    ret void
 //
 //
 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
 // CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {

 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK7:       omp.dispatch.cond:
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK7-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK7:       cond.true:
 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK7-NEXT:    br label [[COND_END:%.*]]
 // CHECK7:       cond.false:
 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    br label [[COND_END]]
 // CHECK7:       cond.end:

 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK8-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK8:       omp.dispatch.cond:
 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK8-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK8:       cond.true:
 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK8-NEXT:    br label [[COND_END:%.*]]
 // CHECK8:       cond.false:
 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    br label [[COND_END]]
 // CHECK8:       cond.end:

 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK13:       omp.dispatch.cond:
 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK13-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK13-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK13:       cond.true:
 // CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK13-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK13-NEXT:    br label [[COND_END:%.*]]
 // CHECK13:       cond.false:
 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK13-NEXT:    br label [[COND_END]]
 // CHECK13:       cond.end:
-// CHECK13-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK13-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK13-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK13:       omp.dispatch.body:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK13:       omp.inner.for.cond:
 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK13-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK13:       omp.inner.for.body:
 // CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
 // CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK13:       omp.body.continue:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK13:       omp.inner.for.inc:
 // CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK13:       omp.inner.for.end:
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK13:       omp.dispatch.inc:
 // CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK13:       omp.dispatch.end:
 // CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK13-NEXT:    ret void
 //
 //
 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
 // CHECK13-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {

 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK14-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK14:       omp.dispatch.cond:
 // CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK14-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK14:       cond.true:
 // CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK14-NEXT:    br label [[COND_END:%.*]]
 // CHECK14:       cond.false:
 // CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK14-NEXT:    br label [[COND_END]]
 // CHECK14:       cond.end:
-// CHECK14-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK14-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK14-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK14:       omp.dispatch.body:
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK14:       omp.inner.for.cond:
 // CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK14-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK14:       omp.inner.for.body:
 // CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
 // CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK14:       omp.body.continue:
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK14:       omp.inner.for.inc:
 // CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK14:       omp.inner.for.end:
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK14:       omp.dispatch.inc:
 // CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK14:       omp.dispatch.end:
 // CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK14-NEXT:    ret void
 //
 //
 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
 // CHECK14-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {

 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK15:       omp.dispatch.cond:
 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK15-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK15:       cond.true:
 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK15-NEXT:    br label [[COND_END:%.*]]
 // CHECK15:       cond.false:
 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    br label [[COND_END]]
 // CHECK15:       cond.end:

 // CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK16-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK16:       omp.dispatch.cond:
 // CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK16-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK16:       cond.true:
 // CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK16-NEXT:    br label [[COND_END:%.*]]
 // CHECK16:       cond.false:
 // CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    br label [[COND_END]]
 // CHECK16:       cond.end:

 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK17:       omp.dispatch.cond:
 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK17-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK17-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK17-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK17:       cond.true:
 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK17-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK17-NEXT:    br label [[COND_END:%.*]]
 // CHECK17:       cond.false:
 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK17-NEXT:    br label [[COND_END]]
 // CHECK17:       cond.end:
-// CHECK17-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK17-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK17-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK17-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK17-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK17:       omp.dispatch.body:
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK17:       omp.inner.for.cond:
 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK17-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK17:       omp.inner.for.body:
 // CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
 // CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK17:       omp.body.continue:
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK17:       omp.inner.for.inc:
 // CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK17:       omp.inner.for.end:
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK17:       omp.dispatch.inc:
 // CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK17-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK17:       omp.dispatch.end:
 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK17-NEXT:    ret void
 //
 //
 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
 // CHECK17-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {

 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK18:       omp.dispatch.cond:
 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK18-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK18-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK18-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK18:       cond.true:
 // CHECK18-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK18-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK18-NEXT:    br label [[COND_END:%.*]]
 // CHECK18:       cond.false:
 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK18-NEXT:    br label [[COND_END]]
 // CHECK18:       cond.end:
-// CHECK18-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK18-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK18-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK18-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK18-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK18:       omp.dispatch.body:
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK18:       omp.inner.for.cond:
 // CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK18-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK18-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK18:       omp.inner.for.body:
 // CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
 // CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
 // CHECK18-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK18-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK18:       omp.body.continue:
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK18:       omp.inner.for.inc:
 // CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK18-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK18:       omp.inner.for.end:
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK18:       omp.dispatch.inc:
 // CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK18-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK18-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK18-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK18:       omp.dispatch.end:
 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK18-NEXT:    ret void
 //
 //
 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
 // CHECK18-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {

 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK19:       omp.dispatch.cond:
 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK19-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK19:       cond.true:
 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK19-NEXT:    br label [[COND_END:%.*]]
 // CHECK19:       cond.false:
 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK19-NEXT:    br label [[COND_END]]
 // CHECK19:       cond.end:

 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK20:       omp.dispatch.cond:
 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK20-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK20:       cond.true:
 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK20-NEXT:    br label [[COND_END:%.*]]
 // CHECK20:       cond.false:
 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK20-NEXT:    br label [[COND_END]]
 // CHECK20:       cond.end:

clang/test/OpenMP/teams_distribute_parallel_for_simd_schedule_codegen.cpp

 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK1-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK1-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    store i32 123, i32* [[I]], align 4

 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK2-NEXT:    br label [[COND_END]]
 // CHECK2:       cond.end:
-// CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK2-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK2-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
-// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
-// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
+// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK2-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK2:       .omp.final.then:
 // CHECK2-NEXT:    store i32 123, i32* [[I]], align 4

 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK3-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK3-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK3-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK3:       omp.dispatch.body:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
-// CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
 // CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
-// CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
+// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK3-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK3:       omp.dispatch.inc:
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK3-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK3:       omp.dispatch.end:
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK3-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    store i32 123, i32* [[I]], align 4

 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK4-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV2]], [[TMP6]]
+// CHECK4-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
+// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK4-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP8]] to i64
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:
-// CHECK4-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP7]], [[COND_TRUE]] ], [ [[CONV3]], [[COND_FALSE]] ]
-// CHECK4-NEXT:    [[CONV4:%.*]] = trunc i64 [[COND]] to i32
-// CHECK4-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
+// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
+// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// CHECK4-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK4-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
+// CHECK4-NEXT:    br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK4:       omp.dispatch.body:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:
 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
-// CHECK4-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
-// CHECK4-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK4-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK4:       omp.inner.for.body:
 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !26
 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !26
 // CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK4-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !26
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK4:       omp.body.continue:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK4:       omp.inner.for.inc:
 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
-// CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
-// CHECK4-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
+// CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK4-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
 // CHECK4:       omp.inner.for.end:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK4:       omp.dispatch.inc:
 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK4-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
+// CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK4-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK4-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK4-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
+// CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK4-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK4:       omp.dispatch.end:
 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK4-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK4:       .omp.final.then:
 // CHECK4-NEXT:    store i32 123, i32* [[I]], align 4

 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK5:       omp.dispatch.cond:
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK5-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK5:       cond.true:
 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK5-NEXT:    br label [[COND_END:%.*]]
 // CHECK5:       cond.false:
 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    br label [[COND_END]]
 // CHECK5:       cond.end:

 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK6-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK6:       omp.dispatch.cond:
 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK6-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK6:       cond.true:
 // CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK6-NEXT:    br label [[COND_END:%.*]]
 // CHECK6:       cond.false:
 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK6-NEXT:    br label [[COND_END]]
 // CHECK6:       cond.end:

 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK7:       omp.dispatch.cond:
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK7-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK7:       cond.true:
 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK7-NEXT:    br label [[COND_END:%.*]]
 // CHECK7:       cond.false:
 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    br label [[COND_END]]
 // CHECK7:       cond.end:

 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
 // CHECK8-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK8:       omp.dispatch.cond:
 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK8-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+// CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
 // CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK8:       cond.true:
 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK8-NEXT:    br label [[COND_END:%.*]]
 // CHECK8:       cond.false:
 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    br label [[COND_END]]
 // CHECK8:       cond.end:

 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK13:       omp.dispatch.cond:
 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK13-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK13-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK13:       cond.true:
 // CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK13-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK13-NEXT:    br label [[COND_END:%.*]]
 // CHECK13:       cond.false:
 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK13-NEXT:    br label [[COND_END]]
 // CHECK13:       cond.end:
-// CHECK13-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK13-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK13-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK13:       omp.dispatch.body:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK13:       omp.inner.for.cond:
 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
-// CHECK13-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK13-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK13-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK13-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK13:       omp.inner.for.body:
 // CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK13:       omp.body.continue:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK13:       omp.inner.for.inc:
 // CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
-// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
+// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
 // CHECK13:       omp.inner.for.end:
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK13:       omp.dispatch.inc:
 // CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK13-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK13-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK13-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK13-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK13-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK13:       omp.dispatch.end:
 // CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK13-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK13-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK13:       .omp.final.then:
 // CHECK13-NEXT:    store i32 10, i32* [[I]], align 4

 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK14-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK14-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK14:       omp.dispatch.cond:
 // CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK14-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK14-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK14-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK14-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK14-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK14:       cond.true:
 // CHECK14-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK14-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK14-NEXT:    br label [[COND_END:%.*]]
 // CHECK14:       cond.false:
 // CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK14-NEXT:    br label [[COND_END]]
 // CHECK14:       cond.end:
-// CHECK14-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK14-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK14-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK14-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK14-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK14-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK14-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK14:       omp.dispatch.body:
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK14:       omp.inner.for.cond:
 // CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
-// CHECK14-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK14-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK14-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK14-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK14:       omp.inner.for.body:
 // CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK14:       omp.body.continue:
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK14:       omp.inner.for.inc:
 // CHECK14-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
-// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
+// CHECK14-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK14-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
 // CHECK14:       omp.inner.for.end:
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK14:       omp.dispatch.inc:
 // CHECK14-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK14-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK14-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK14-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK14-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK14-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK14-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK14-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK14-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK14:       omp.dispatch.end:
 // CHECK14-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK14-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK14-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK14-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK14:       .omp.final.then:
 // CHECK14-NEXT:    store i32 10, i32* [[I]], align 4

 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK15-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK15-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK15:       omp.dispatch.cond:
 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK15-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK15-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK15:       cond.true:
 // CHECK15-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK15-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK15-NEXT:    br label [[COND_END:%.*]]
 // CHECK15:       cond.false:
 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK15-NEXT:    br label [[COND_END]]
 // CHECK15:       cond.end:
-// CHECK15-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK15-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK15-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK15-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK15-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK15-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK15-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK15-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK15:       omp.dispatch.body:
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK15:       omp.inner.for.cond:
 // CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
-// CHECK15-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK15-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK15-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK15-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK15:       omp.inner.for.body:
 // CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
 // CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
 // CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK15:       omp.body.continue:
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK15:       omp.inner.for.inc:
 // CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
-// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
+// CHECK15-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK15-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
 // CHECK15:       omp.inner.for.end:
 // CHECK15-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK15:       omp.dispatch.inc:
 // CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK15-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK15-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK15-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK15-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK15-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK15-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK15-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK15-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK15-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK15:       omp.dispatch.end:
 // CHECK15-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK15-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK15-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK15-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK15:       .omp.final.then:
 // CHECK15-NEXT:    store i32 10, i32* [[I]], align 4

 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
 // CHECK16-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK16-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK16-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK16:       omp.dispatch.cond:
 // CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK16-NEXT:    [[CONV3:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK16-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK16-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[CONV3]], [[TMP7]]
+// CHECK16-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
+// CHECK16-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
 // CHECK16-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK16:       cond.true:
 // CHECK16-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK16-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK16-NEXT:    br label [[COND_END:%.*]]
 // CHECK16:       cond.false:
 // CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK16-NEXT:    [[CONV4:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK16-NEXT:    br label [[COND_END]]
 // CHECK16:       cond.end:
-// CHECK16-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP8]], [[COND_TRUE]] ], [ [[CONV4]], [[COND_FALSE]] ]
-// CHECK16-NEXT:    [[CONV5:%.*]] = trunc i64 [[COND]] to i32
-// CHECK16-NEXT:    store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
+// CHECK16-NEXT:    [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK16-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK16-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
 // CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK16-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
-// CHECK16-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK16-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK16-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK16:       omp.dispatch.body:
 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK16:       omp.inner.for.cond:
 // CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !61
-// CHECK16-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
-// CHECK16-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK16-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK16-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK16:       omp.inner.for.body:
 // CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !61
 // CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !61
 // CHECK16-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !61
 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK16:       omp.body.continue:
 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK16:       omp.inner.for.inc:
 // CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
-// CHECK16-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], 1
-// CHECK16-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
+// CHECK16-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK16-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !61
 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP62:![0-9]+]]
 // CHECK16:       omp.inner.for.end:
 // CHECK16-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK16:       omp.dispatch.inc:
 // CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK16-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK16-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
+// CHECK16-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
+// CHECK16-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
 // CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
-// CHECK16-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK16-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
+// CHECK16-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
+// CHECK16-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
 // CHECK16-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK16:       omp.dispatch.end:
 // CHECK16-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
 // CHECK16-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK16-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK16-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK16:       .omp.final.then:
 // CHECK16-NEXT:    store i32 10, i32* [[I]], align 4

 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK17:       omp.dispatch.cond:
 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK17-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK17:       cond.true:
 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK17-NEXT:    br label [[COND_END:%.*]]
 // CHECK17:       cond.false:
 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK17-NEXT:    br label [[COND_END]]
 // CHECK17:       cond.end:

 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK18:       omp.dispatch.cond:
 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK18-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK18:       cond.true:
 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK18-NEXT:    br label [[COND_END:%.*]]
 // CHECK18:       cond.false:
 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK18-NEXT:    br label [[COND_END]]
 // CHECK18:       cond.end:

 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK19:       omp.dispatch.cond:
 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK19-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK19:       cond.true:
 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK19-NEXT:    br label [[COND_END:%.*]]
 // CHECK19:       cond.false:
 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK19-NEXT:    br label [[COND_END]]
 // CHECK19:       cond.end:

 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
 // CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK20:       omp.dispatch.cond:
 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK20-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP6]], [[TMP7]]
+// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK20:       cond.true:
 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK20-NEXT:    br label [[COND_END:%.*]]
 // CHECK20:       cond.false:
 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK20-NEXT:    br label [[COND_END]]
 // CHECK20:       cond.end: