[OPENMP]Fix PR50129: omp cancel parallel not working as expected.

[ABataev]

Summary

Need to emit a call for kmpc_cancel_barrier in the exit block for
kmpc_cancel function call.

[jdoerfert]

Do we have to update OpenMPIRBuilder::createCancel too?

[ABataev]

1. Emit __kmpc_cancel_barrier only for parallel cancellation.
2. Synced CreateCancel.

[jdoerfert]

LG, one nit below.

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

537

The documentation doesn't match the prototype.

clang/lib/CodeGen/CGOpenMPRuntime.cpp

if (CancelRegion == OMPD_taskgroup || OMPRegionInfo->hasCancel()) {
           emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
           CGF.Builder.getInt32(getCancellationKind(CancelRegion))};
       // Ignore return result until untied tasks are supported.
       llvm::Value *Result = CGF.EmitRuntimeCall(
           OMPBuilder.getOrCreateRuntimeFunction(
               CGM.getModule(), OMPRTL___kmpc_cancellationpoint),
           Args);
       // if (__kmpc_cancellationpoint()) {
+      //   call i32 @__kmpc_cancel_barrier( // for parallel cancellation only
       //   exit from construct;
       // }
       llvm::BasicBlock *ExitBB = CGF.createBasicBlock(".cancel.exit");
       llvm::BasicBlock *ContBB = CGF.createBasicBlock(".cancel.continue");
       llvm::Value *Cmp = CGF.Builder.CreateIsNotNull(Result);
       CGF.Builder.CreateCondBr(Cmp, ExitBB, ContBB);
       CGF.EmitBlock(ExitBB);
+      if (CancelRegion == OMPD_parallel)
+        emitBarrierCall(CGF, Loc, OMPD_unknown, /*EmitChecks=*/false);
       // exit from construct;
       CodeGenFunction::JumpDest CancelDest =
           CGF.getOMPCancelDestination(OMPRegionInfo->getDirectiveKind());
       CGF.EmitBranchThroughCleanup(CancelDest);
       CGF.EmitBlock(ContBB, /*IsFinished=*/true);
     }
   }
 }
auto &&ThenGen = [this, &M, Loc, CancelRegion,
       CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
       llvm::Value *Args[] = {
           RT.emitUpdateLocation(CGF, Loc), RT.getThreadID(CGF, Loc),
           CGF.Builder.getInt32(getCancellationKind(CancelRegion))};
       // Ignore return result until untied tasks are supported.
       llvm::Value *Result = CGF.EmitRuntimeCall(
           OMPBuilder.getOrCreateRuntimeFunction(M, OMPRTL___kmpc_cancel), Args);
       // if (__kmpc_cancel()) {
+      //   call i32 @__kmpc_cancel_barrier( // for parallel cancellation only
       //   exit from construct;
       // }
       llvm::BasicBlock *ExitBB = CGF.createBasicBlock(".cancel.exit");
       llvm::BasicBlock *ContBB = CGF.createBasicBlock(".cancel.continue");
       llvm::Value *Cmp = CGF.Builder.CreateIsNotNull(Result);
       CGF.Builder.CreateCondBr(Cmp, ExitBB, ContBB);
       CGF.EmitBlock(ExitBB);
+      if (CancelRegion == OMPD_parallel)
+        RT.emitBarrierCall(CGF, Loc, OMPD_unknown, /*EmitChecks=*/false);
       // exit from construct;
       CodeGenFunction::JumpDest CancelDest =
           CGF.getOMPCancelDestination(OMPRegionInfo->getDirectiveKind());
       CGF.EmitBranchThroughCleanup(CancelDest);
       CGF.EmitBlock(ContBB, /*IsFinished=*/true);
     };
     if (IfCond) {
       emitIfClause(CGF, IfCond, ThenGen,

clang/test/OpenMP/cancel_codegen.cpp

 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK1-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK1-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
 // CHECK1-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:

 // CHECK1:       .omp.sections.exit:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK1:       cancel.cont:
-// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
 // CHECK1-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
 // CHECK1-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
 // CHECK1:       omp.inner.for.cond6:

 // CHECK1-NEXT:    ]
 // CHECK1:       .omp.sections.case9:
 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK1-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
 // CHECK1:       .cancel.exit10:
 // CHECK1-NEXT:    br label [[CANCEL_EXIT19:%.*]]
 // CHECK1:       cancel.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
 // CHECK1:       .cancel.continue11:
 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
 // CHECK1:       .omp.sections.case12:
 // CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK1-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK1-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
 // CHECK1:       .cancel.exit13:
 // CHECK1-NEXT:    br label [[CANCEL_EXIT19]]
 // CHECK1:       .cancel.continue14:
 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
 // CHECK1:       .omp.sections.exit15:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
 // CHECK1:       omp.inner.for.inc16:
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK1-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK1-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND6]]
 // CHECK1:       omp.inner.for.end18:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT20:%.*]]
 // CHECK1:       cancel.cont20:
-// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK1-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK1-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
 // CHECK1-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK1-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK1-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
 // CHECK1-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:

 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK1:       omp_if.then:
 // CHECK1-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
 // CHECK1-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
 // CHECK1-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
 // CHECK1:       .cancel.exit29:
 // CHECK1-NEXT:    br label [[CANCEL_EXIT34:%.*]]
 // CHECK1:       cancel.exit19:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT20]]
 // CHECK1:       .cancel.continue30:
 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK1:       omp_if.else:
 // CHECK1-NEXT:    br label [[OMP_IF_END]]
 // CHECK1:       omp_if.end:
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
 // CHECK1:       omp.inner.for.inc31:
 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
 // CHECK1-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND26]]
 // CHECK1:       omp.inner.for.end33:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       cancel.exit34:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT35:%.*]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    br label [[CANCEL_CONT35]]
 // CHECK1:       cancel.cont35:
-// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK1-NEXT:    [[TMP39:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK1-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*))
 // CHECK1-NEXT:    store i32 0, i32* [[R]], align 4
 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])

 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK1:       omp_if.then:
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
 // CHECK1-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
 // CHECK1-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK1:       .cancel.exit:
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]])
 // CHECK1-NEXT:    br label [[RETURN:%.*]]
 // CHECK1:       .cancel.continue:
 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK1:       omp_if.else:
 // CHECK1-NEXT:    br label [[OMP_IF_END]]
 // CHECK1:       omp_if.end:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP7]] to i8
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 0
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 0
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
+// CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK1-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP11]])
-// CHECK1-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
-// CHECK1-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+// CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP14]])
+// CHECK1-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
+// CHECK1-NEXT:    br i1 [[TMP16]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
 // CHECK1:       .cancel.exit2:
 // CHECK1-NEXT:    br label [[RETURN]]
 // CHECK1:       .cancel.continue3:
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK1-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP15]], i64 0
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
-// CHECK1-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP16]], i64 0
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
-// CHECK1-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP17]] to i32
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP14]]
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP18]], i64 0
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
+// CHECK1-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP19]], i64 0
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
+// CHECK1-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP20]] to i32
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP17]]
 // CHECK1-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
 // CHECK1-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
 // CHECK1-NEXT:    br label [[RETURN]]
 // CHECK1:       return:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.

 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:

 // CHECK1:       .omp.sections.exit:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK1:       cancel.cont:
 // CHECK1-NEXT:    ret void
 // CHECK1:       cancel.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
 // CHECK1-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:

 // CHECK1:       .omp.sections.exit:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK1:       cancel.cont:
 // CHECK1-NEXT:    ret void
 // CHECK1:       cancel.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8

 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[R3]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
 // CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:

 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP24]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP24]])
 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
 // CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK1-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB7:[0-9]+]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
 // CHECK1-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK1-NEXT:    ]
 // CHECK1:       .omp.reduction.case1:
 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
 // CHECK1-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK1:       cancel.exit:
 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP34]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP34]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK1:       .omp.reduction.case2:
 // CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
 // CHECK1-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK1:       .omp.reduction.default:
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:

 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
 // CHECK2-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:

 // CHECK2:       .omp.sections.exit:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK2:       cancel.cont:
-// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
 // CHECK2-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
 // CHECK2-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
 // CHECK2:       omp.inner.for.cond6:

 // CHECK2-NEXT:    ]
 // CHECK2:       .omp.sections.case9:
 // CHECK2-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK2-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK2-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
 // CHECK2:       .cancel.exit10:
 // CHECK2-NEXT:    br label [[CANCEL_EXIT19:%.*]]
 // CHECK2:       cancel.exit:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
 // CHECK2:       .cancel.continue11:
 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
 // CHECK2:       .omp.sections.case12:
 // CHECK2-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK2-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK2-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
 // CHECK2:       .cancel.exit13:
 // CHECK2-NEXT:    br label [[CANCEL_EXIT19]]
 // CHECK2:       .cancel.continue14:
 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
 // CHECK2:       .omp.sections.exit15:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
 // CHECK2:       omp.inner.for.inc16:
 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK2-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK2-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND6]]
 // CHECK2:       omp.inner.for.end18:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT20:%.*]]
 // CHECK2:       cancel.cont20:
-// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK2-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK2-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK2-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
 // CHECK2-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK2:       omp.precond.then:
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK2-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], 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:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK2-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
 // CHECK2-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:

 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK2:       omp_if.then:
 // CHECK2-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
 // CHECK2-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
 // CHECK2-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
 // CHECK2:       .cancel.exit29:
 // CHECK2-NEXT:    br label [[CANCEL_EXIT34:%.*]]
 // CHECK2:       cancel.exit19:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT20]]
 // CHECK2:       .cancel.continue30:
 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK2:       omp_if.else:
 // CHECK2-NEXT:    br label [[OMP_IF_END]]
 // CHECK2:       omp_if.end:
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
 // CHECK2:       omp.inner.for.inc31:
 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
 // CHECK2-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND26]]
 // CHECK2:       omp.inner.for.end33:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       cancel.exit34:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT35:%.*]]
 // CHECK2:       omp.precond.end:
 // CHECK2-NEXT:    br label [[CANCEL_CONT35]]
 // CHECK2:       cancel.cont35:
-// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK2-NEXT:    [[TMP39:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK2-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*))
 // CHECK2-NEXT:    store i32 0, i32* [[R]], align 4
 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])

 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK2:       omp_if.then:
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
 // CHECK2-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
 // CHECK2-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK2:       .cancel.exit:
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]])
 // CHECK2-NEXT:    br label [[RETURN:%.*]]
 // CHECK2:       .cancel.continue:
 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK2:       omp_if.else:
 // CHECK2-NEXT:    br label [[OMP_IF_END]]
 // CHECK2:       omp_if.end:
-// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP7]] to i8
-// CHECK2-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 0
-// CHECK2-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 0
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
+// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK2-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
-// CHECK2-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
-// CHECK2-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP11]])
-// CHECK2-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
-// CHECK2-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+// CHECK2-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP14]])
+// CHECK2-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
+// CHECK2-NEXT:    br i1 [[TMP16]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
 // CHECK2:       .cancel.exit2:
 // CHECK2-NEXT:    br label [[RETURN]]
 // CHECK2:       .cancel.continue3:
-// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK2-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP15]], i64 0
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
-// CHECK2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP16]], i64 0
-// CHECK2-NEXT:    [[TMP17:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
-// CHECK2-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP17]] to i32
-// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP14]]
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP18]], i64 0
+// CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
+// CHECK2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP19]], i64 0
+// CHECK2-NEXT:    [[TMP20:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
+// CHECK2-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP20]] to i32
+// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP17]]
 // CHECK2-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
 // CHECK2-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
 // CHECK2-NEXT:    br label [[RETURN]]
 // CHECK2:       return:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.

 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
 // CHECK2-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:

 // CHECK2:       .omp.sections.exit:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK2:       cancel.cont:
 // CHECK2-NEXT:    ret void
 // CHECK2:       cancel.exit:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
 // CHECK2-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:

 // CHECK2:       .omp.sections.exit:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK2:       cancel.cont:
 // CHECK2-NEXT:    ret void
 // CHECK2:       cancel.exit:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8

 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], 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:    store i32 0, i32* [[R3]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
 // CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:

 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP24]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP24]])
 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 // CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
 // CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
 // CHECK2-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK2-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB7:[0-9]+]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
 // CHECK2-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK2-NEXT:    ]
 // CHECK2:       .omp.reduction.case1:
 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
 // CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
 // CHECK2-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK2:       cancel.exit:
 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP34]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP34]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK2:       .omp.reduction.case2:
 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
 // CHECK2-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK2:       .omp.reduction.default:
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       omp.precond.end:

 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
 // CHECK3-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_LASTITER26:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_LOWERBOUND27:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_UPPERBOUND28:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_STRIDE29:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_33:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I35:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[I36:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
 // CHECK3-NEXT:    [[R:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK3-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK3-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
 // CHECK3-NEXT:    br label [[OMP_PARALLEL:%.*]]
 // CHECK3:       omp_parallel:
 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @main..omp_par to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
 // CHECK3:       omp.par.outlined.exit:
 // CHECK3-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
 // CHECK3:       omp.par.exit.split:
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
 // CHECK3:       omp_section_loop.preheader:
 // CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
 // CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
 // CHECK3-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
 // CHECK3:       omp_section_loop.header:
 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
 // CHECK3:       omp_section_loop.cond:
 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
 // CHECK3:       omp_section_loop.body:
 // CHECK3-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
 // CHECK3-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
 // CHECK3-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
 // CHECK3-NEXT:    ]
 // CHECK3:       omp_section_loop.inc:
 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
 // CHECK3:       omp_section_loop.exit:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
 // CHECK3:       omp_section_loop.after:
 // CHECK3-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
 // CHECK3:       omp_sections.end:
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER12:%.*]]
-// CHECK3:       omp_section_loop.preheader12:
-// CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND27]], align 4
-// CHECK3-NEXT:    store i32 1, i32* [[P_UPPERBOUND28]], align 4
-// CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE29]], align 4
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM30:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]], i32 34, i32* [[P_LASTITER26]], i32* [[P_LOWERBOUND27]], i32* [[P_UPPERBOUND28]], i32* [[P_STRIDE29]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND27]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND28]], align 4
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
+// CHECK3:       omp_section_loop.preheader13:
+// CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
+// CHECK3-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
+// CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 1)
+// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
 // CHECK3-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13:%.*]]
-// CHECK3:       omp_section_loop.header13:
-// CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV19:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER12]] ], [ [[OMP_SECTION_LOOP_NEXT21:%.*]], [[OMP_SECTION_LOOP_INC16:%.*]] ]
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND14:%.*]]
-// CHECK3:       omp_section_loop.cond14:
-// CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP20:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV19]], [[TMP10]]
-// CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP20]], label [[OMP_SECTION_LOOP_BODY15:%.*]], label [[OMP_SECTION_LOOP_EXIT17:%.*]]
-// CHECK3:       omp_section_loop.body15:
-// CHECK3-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV19]], [[TMP7]]
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
+// CHECK3:       omp_section_loop.header14:
+// CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
+// CHECK3:       omp_section_loop.cond15:
+// CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
+// CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
+// CHECK3:       omp_section_loop.body16:
+// CHECK3-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
 // CHECK3-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
 // CHECK3-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
-// CHECK3-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC16]] [
-// CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE22:%.*]]
-// CHECK3-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE24:%.*]]
+// CHECK3-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
+// CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
+// CHECK3-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
 // CHECK3-NEXT:    ]
-// CHECK3:       omp_section_loop.inc16:
-// CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT21]] = add nuw i32 [[OMP_SECTION_LOOP_IV19]], 1
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13]]
-// CHECK3:       omp_section_loop.exit17:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]])
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER18:%.*]]
-// CHECK3:       omp_section_loop.after18:
-// CHECK3-NEXT:    br label [[OMP_SECTIONS_END32:%.*]]
-// CHECK3:       omp_sections.end32:
+// CHECK3:       omp_section_loop.inc17:
+// CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
+// CHECK3:       omp_section_loop.exit18:
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
+// CHECK3:       omp_section_loop.after19:
+// CHECK3-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
+// CHECK3:       omp_sections.end33:
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK3-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK3-NEXT:    store i32 [[SUB34]], i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK3-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK3-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK3:       omp.precond.then:
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK3-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], 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:    [[OMP_GLOBAL_THREAD_NUM36:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM36]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
-// CHECK3-NEXT:    [[CMP37:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
-// CHECK3-NEXT:    br i1 [[CMP37]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK3-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
+// CHECK3-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
-// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
-// CHECK3-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
+// CHECK3-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
-// CHECK3-NEXT:    [[ADD39:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK3-NEXT:    store i32 [[ADD39]], i32* [[I35]], align 4
+// CHECK3-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK3-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
 // CHECK3-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
-// CHECK3-NEXT:    [[TOBOOL40:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
-// CHECK3-NEXT:    br i1 [[TOBOOL40]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
+// CHECK3-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+// CHECK3-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK3:       omp_if.then:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
-// CHECK3-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]], i32 2)
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
+// CHECK3-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
 // CHECK3-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
 // CHECK3-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK3:       .cancel.exit:
 // CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
 // CHECK3:       omp_section_loop.body.case:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM9]], i32 3)
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
 // CHECK3-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
 // CHECK3-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
 // CHECK3:       omp_section_loop.body.case.split:
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
 // CHECK3:       omp_section_loop.body.case.cncl:
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
-// CHECK3:       omp_section_loop.body.case22:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]], i32 3)
+// CHECK3:       omp_section_loop.body.case23:
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
 // CHECK3-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
-// CHECK3-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE22_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE22_CNCL:%.*]]
-// CHECK3:       omp_section_loop.body.case22.split:
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK3:       omp_section_loop.body.case22.cncl:
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK3:       omp_section_loop.body.case24:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM25:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM25]], i32 3)
+// CHECK3-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
+// CHECK3:       omp_section_loop.body.case23.split:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK3:       omp_section_loop.body.case23.cncl:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK3:       omp_section_loop.body.case25:
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
 // CHECK3-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
-// CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE24_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE24_CNCL:%.*]]
-// CHECK3:       omp_section_loop.body.case24.split:
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK3:       omp_section_loop.body.case24.cncl:
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
+// CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
+// CHECK3:       omp_section_loop.body.case25.split:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK3:       omp_section_loop.body.case25.cncl:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK3:       .cancel.continue:
 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK3:       omp_if.else:
 // CHECK3-NEXT:    br label [[OMP_IF_END]]
 // CHECK3:       omp_if.end:
 // 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:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP35]], 1
-// CHECK3-NEXT:    store i32 [[ADD42]], i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
+// CHECK3-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       cancel.exit:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM43:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM43]])
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK3:       omp.precond.end:
 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
 // CHECK3:       cancel.cont:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
-// CHECK3-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM46]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
+// CHECK3-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
-// CHECK3-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i8* [[TMP36]])
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
+// CHECK3-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
 // CHECK3-NEXT:    store i32 0, i32* [[R]], align 4
 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
 // CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK3-NEXT:    ret i32 [[TMP40]]
 //
 //

 // CHECK3:       2:
 // CHECK3-NEXT:    br label [[TMP3:%.*]]
 // CHECK3:       3:
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
-// CHECK3-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX2]], align 1
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK3-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
+// CHECK3-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
-// CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL4:%.*]]
-// CHECK3:       .cncl4:
+// CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
+// CHECK3:       .cncl5:
 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
 // CHECK3:       .cont:
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
-// CHECK3-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX5]], align 8
-// CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX6]], align 1
-// CHECK3-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP12]] to i32
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV7]], [[TMP9]]
-// CHECK3-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD]] to i8
-// CHECK3-NEXT:    store i8 [[CONV8]], i8* [[ARRAYIDX6]], align 1
+// CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
+// CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
+// CHECK3-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP9]]
+// CHECK3-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
+// CHECK3-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
 // CHECK3-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
 // CHECK3:       omp.par.pre_finalize:
 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK3:       13:
 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
 // CHECK3:       .cncl:
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK3:       .split:
 // CHECK3-NEXT:    br label [[TMP3]]
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
 // CHECK3-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK3-NEXT:  entry:

 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
 // CHECK4-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_LASTITER26:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_LOWERBOUND27:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_UPPERBOUND28:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_STRIDE29:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_33:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[I35:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[I36:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
 // CHECK4-NEXT:    [[R:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK4-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK4-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
 // CHECK4-NEXT:    br label [[OMP_PARALLEL:%.*]]
 // CHECK4:       omp_parallel:
 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @main..omp_par to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
 // CHECK4:       omp.par.outlined.exit:
 // CHECK4-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
 // CHECK4:       omp.par.exit.split:
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
 // CHECK4:       omp_section_loop.preheader:
 // CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
 // CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
 // CHECK4-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
 // CHECK4-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
 // CHECK4:       omp_section_loop.header:
 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
 // CHECK4:       omp_section_loop.cond:
 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
 // CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
 // CHECK4:       omp_section_loop.body:
 // CHECK4-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
 // CHECK4-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
 // CHECK4-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
 // CHECK4-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
 // CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
 // CHECK4-NEXT:    ]
 // CHECK4:       omp_section_loop.inc:
 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
 // CHECK4:       omp_section_loop.exit:
-// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
 // CHECK4:       omp_section_loop.after:
 // CHECK4-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
 // CHECK4:       omp_sections.end:
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER12:%.*]]
-// CHECK4:       omp_section_loop.preheader12:
-// CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND27]], align 4
-// CHECK4-NEXT:    store i32 1, i32* [[P_UPPERBOUND28]], align 4
-// CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE29]], align 4
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM30:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]], i32 34, i32* [[P_LASTITER26]], i32* [[P_LOWERBOUND27]], i32* [[P_UPPERBOUND28]], i32* [[P_STRIDE29]], i32 1, i32 1)
-// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND27]], align 4
-// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND28]], align 4
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
+// CHECK4:       omp_section_loop.preheader13:
+// CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
+// CHECK4-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
+// CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 1)
+// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
+// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
 // CHECK4-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
 // CHECK4-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13:%.*]]
-// CHECK4:       omp_section_loop.header13:
-// CHECK4-NEXT:    [[OMP_SECTION_LOOP_IV19:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER12]] ], [ [[OMP_SECTION_LOOP_NEXT21:%.*]], [[OMP_SECTION_LOOP_INC16:%.*]] ]
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND14:%.*]]
-// CHECK4:       omp_section_loop.cond14:
-// CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP20:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV19]], [[TMP10]]
-// CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP20]], label [[OMP_SECTION_LOOP_BODY15:%.*]], label [[OMP_SECTION_LOOP_EXIT17:%.*]]
-// CHECK4:       omp_section_loop.body15:
-// CHECK4-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV19]], [[TMP7]]
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
+// CHECK4:       omp_section_loop.header14:
+// CHECK4-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
+// CHECK4:       omp_section_loop.cond15:
+// CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
+// CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
+// CHECK4:       omp_section_loop.body16:
+// CHECK4-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
 // CHECK4-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
 // CHECK4-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
-// CHECK4-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC16]] [
-// CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE22:%.*]]
-// CHECK4-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE24:%.*]]
+// CHECK4-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
+// CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
+// CHECK4-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
 // CHECK4-NEXT:    ]
-// CHECK4:       omp_section_loop.inc16:
-// CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT21]] = add nuw i32 [[OMP_SECTION_LOOP_IV19]], 1
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13]]
-// CHECK4:       omp_section_loop.exit17:
-// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]])
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER18:%.*]]
-// CHECK4:       omp_section_loop.after18:
-// CHECK4-NEXT:    br label [[OMP_SECTIONS_END32:%.*]]
-// CHECK4:       omp_sections.end32:
+// CHECK4:       omp_section_loop.inc17:
+// CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
+// CHECK4:       omp_section_loop.exit18:
+// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
+// CHECK4:       omp_section_loop.after19:
+// CHECK4-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
+// CHECK4:       omp_sections.end33:
 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
 // CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK4-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK4-NEXT:    store i32 [[SUB34]], i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK4-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK4-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK4:       omp.precond.then:
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
-// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK4-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], 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:    [[OMP_GLOBAL_THREAD_NUM36:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
-// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM36]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
+// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
-// CHECK4-NEXT:    [[CMP37:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
-// CHECK4-NEXT:    br i1 [[CMP37]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK4-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
+// CHECK4-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
-// CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:
 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:
 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
-// CHECK4-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK4-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
+// CHECK4-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK4:       omp.inner.for.body:
 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
-// CHECK4-NEXT:    [[ADD39:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK4-NEXT:    store i32 [[ADD39]], i32* [[I35]], align 4
+// CHECK4-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK4-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
 // CHECK4-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
-// CHECK4-NEXT:    [[TOBOOL40:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
-// CHECK4-NEXT:    br i1 [[TOBOOL40]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
+// CHECK4-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+// CHECK4-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK4:       omp_if.then:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
-// CHECK4-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]], i32 2)
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
+// CHECK4-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
 // CHECK4-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
 // CHECK4-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK4:       .cancel.exit:
 // CHECK4-NEXT:    br label [[CANCEL_EXIT:%.*]]
 // CHECK4:       omp_section_loop.body.case:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM9]], i32 3)
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
 // CHECK4-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
 // CHECK4-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
 // CHECK4:       omp_section_loop.body.case.split:
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
 // CHECK4:       omp_section_loop.body.case.cncl:
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
-// CHECK4:       omp_section_loop.body.case22:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]], i32 3)
+// CHECK4:       omp_section_loop.body.case23:
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
 // CHECK4-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
-// CHECK4-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE22_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE22_CNCL:%.*]]
-// CHECK4:       omp_section_loop.body.case22.split:
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK4:       omp_section_loop.body.case22.cncl:
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK4:       omp_section_loop.body.case24:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM25:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM25]], i32 3)
+// CHECK4-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
+// CHECK4:       omp_section_loop.body.case23.split:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK4:       omp_section_loop.body.case23.cncl:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK4:       omp_section_loop.body.case25:
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
 // CHECK4-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
-// CHECK4-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE24_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE24_CNCL:%.*]]
-// CHECK4:       omp_section_loop.body.case24.split:
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK4:       omp_section_loop.body.case24.cncl:
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
+// CHECK4-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
+// CHECK4:       omp_section_loop.body.case25.split:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK4:       omp_section_loop.body.case25.cncl:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK4:       .cancel.continue:
 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK4:       omp_if.else:
 // CHECK4-NEXT:    br label [[OMP_IF_END]]
 // CHECK4:       omp_if.end:
 // 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:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK4-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP35]], 1
-// CHECK4-NEXT:    store i32 [[ADD42]], i32* [[DOTOMP_IV]], align 4
+// CHECK4-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
+// CHECK4-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK4:       omp.inner.for.end:
 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK4:       omp.loop.exit:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
-// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
+// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK4:       cancel.exit:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM43:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
-// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM43]])
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK4:       omp.precond.end:
 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
 // CHECK4:       cancel.cont:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
-// CHECK4-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM46]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
+// CHECK4-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
-// CHECK4-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i8* [[TMP36]])
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
+// CHECK4-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
 // CHECK4-NEXT:    store i32 0, i32* [[R]], align 4
 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
 // CHECK4-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK4-NEXT:    ret i32 [[TMP40]]
 //
 //

 // CHECK4:       2:
 // CHECK4-NEXT:    br label [[TMP3:%.*]]
 // CHECK4:       3:
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
-// CHECK4-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX2]], align 1
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK4-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
+// CHECK4-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
 // CHECK4-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
-// CHECK4-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL4:%.*]]
-// CHECK4:       .cncl4:
+// CHECK4-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
+// CHECK4:       .cncl5:
 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
 // CHECK4:       .cont:
 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK4-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
-// CHECK4-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
-// CHECK4-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX5]], align 8
-// CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
-// CHECK4-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX6]], align 1
-// CHECK4-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP12]] to i32
-// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV7]], [[TMP9]]
-// CHECK4-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD]] to i8
-// CHECK4-NEXT:    store i8 [[CONV8]], i8* [[ARRAYIDX6]], align 1
+// CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
+// CHECK4-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
+// CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
+// CHECK4-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
+// CHECK4-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP9]]
+// CHECK4-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
+// CHECK4-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
 // CHECK4-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
 // CHECK4:       omp.par.pre_finalize:
 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK4:       13:
 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
 // CHECK4-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
 // CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
 // CHECK4:       .cncl:
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK4:       .split:
 // CHECK4-NEXT:    br label [[TMP3]]
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@.omp_task_entry.
 // CHECK4-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK4-NEXT:  entry:

 // CHECK7-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK7-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK7-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
-// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
 // CHECK7-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK7:       omp.inner.for.cond:

 // CHECK7:       .omp.sections.exit:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK7:       omp.inner.for.inc:
 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK7:       omp.inner.for.end:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK7:       cancel.cont:
-// CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
-// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
+// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK7-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
 // CHECK7-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
 // CHECK7-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
 // CHECK7-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
 // CHECK7:       omp.inner.for.cond6:

 // CHECK7-NEXT:    ]
 // CHECK7:       .omp.sections.case9:
 // CHECK7-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK7-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK7-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
 // CHECK7:       .cancel.exit10:
 // CHECK7-NEXT:    br label [[CANCEL_EXIT19:%.*]]
 // CHECK7:       cancel.exit:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
 // CHECK7:       .cancel.continue11:
 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
 // CHECK7:       .omp.sections.case12:
 // CHECK7-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK7-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK7-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
 // CHECK7:       .cancel.exit13:
 // CHECK7-NEXT:    br label [[CANCEL_EXIT19]]
 // CHECK7:       .cancel.continue14:
 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
 // CHECK7:       .omp.sections.exit15:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
 // CHECK7:       omp.inner.for.inc16:
 // CHECK7-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK7-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK7-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND6]]
 // CHECK7:       omp.inner.for.end18:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT20:%.*]]
 // CHECK7:       cancel.cont20:
-// CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
 // CHECK7-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK7-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK7-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK7-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
 // CHECK7-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK7-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK7-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK7-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK7-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
 // CHECK7-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK7:       omp.precond.then:
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK7-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK7-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK7-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
 // CHECK7-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK7:       cond.true:
 // CHECK7-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK7-NEXT:    br label [[COND_END:%.*]]
 // CHECK7:       cond.false:

 // CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK7:       omp_if.then:
 // CHECK7-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
 // CHECK7-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
 // CHECK7-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
 // CHECK7:       .cancel.exit29:
 // CHECK7-NEXT:    br label [[CANCEL_EXIT34:%.*]]
 // CHECK7:       cancel.exit19:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT20]]
 // CHECK7:       .cancel.continue30:
 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK7:       omp_if.else:
 // CHECK7-NEXT:    br label [[OMP_IF_END]]
 // CHECK7:       omp_if.end:
 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK7:       omp.body.continue:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
 // CHECK7:       omp.inner.for.inc31:
 // CHECK7-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
 // CHECK7-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND26]]
 // CHECK7:       omp.inner.for.end33:
 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK7:       omp.loop.exit:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
 // CHECK7-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK7:       cancel.exit34:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT35:%.*]]
 // CHECK7:       omp.precond.end:
 // CHECK7-NEXT:    br label [[CANCEL_CONT35]]
 // CHECK7:       cancel.cont35:
-// CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]])
+// CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK7-NEXT:    [[TMP39:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK7-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
 // CHECK7-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
 // CHECK7-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*))
 // CHECK7-NEXT:    store i32 0, i32* [[R]], align 4
 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])

 // CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK7:       omp_if.then:
 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK7-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
 // CHECK7-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
 // CHECK7-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK7:       .cancel.exit:
+// CHECK7-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK7-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]])
 // CHECK7-NEXT:    br label [[RETURN:%.*]]
 // CHECK7:       .cancel.continue:
 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK7:       omp_if.else:
 // CHECK7-NEXT:    br label [[OMP_IF_END]]
 // CHECK7:       omp_if.end:
-// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK7-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP7]] to i8
-// CHECK7-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 0
-// CHECK7-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK7-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 0
+// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK7-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
+// CHECK7-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
+// CHECK7-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
+// CHECK7-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK7-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
-// CHECK7-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
-// CHECK7-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP11]])
-// CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
-// CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
+// CHECK7-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+// CHECK7-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP14]])
+// CHECK7-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
+// CHECK7-NEXT:    br i1 [[TMP16]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
 // CHECK7:       .cancel.exit2:
 // CHECK7-NEXT:    br label [[RETURN]]
 // CHECK7:       .cancel.continue3:
-// CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK7-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK7-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP15]], i64 0
-// CHECK7-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
-// CHECK7-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP16]], i64 0
-// CHECK7-NEXT:    [[TMP17:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
-// CHECK7-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP17]] to i32
-// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP14]]
+// CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK7-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK7-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP18]], i64 0
+// CHECK7-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
+// CHECK7-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP19]], i64 0
+// CHECK7-NEXT:    [[TMP20:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
+// CHECK7-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP20]] to i32
+// CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP17]]
 // CHECK7-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
 // CHECK7-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
 // CHECK7-NEXT:    br label [[RETURN]]
 // CHECK7:       return:
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@.omp_task_entry.

 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
 // CHECK7-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK7:       omp.inner.for.cond:

 // CHECK7:       .omp.sections.exit:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK7:       omp.inner.for.inc:
 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK7:       omp.inner.for.end:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK7:       cancel.cont:
 // CHECK7-NEXT:    ret void
 // CHECK7:       cancel.exit:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..3
 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
 // CHECK7-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK7:       omp.inner.for.cond:

 // CHECK7:       .omp.sections.exit:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK7:       omp.inner.for.inc:
 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK7:       omp.inner.for.end:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK7:       cancel.cont:
 // CHECK7-NEXT:    ret void
 // CHECK7:       cancel.exit:
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..4
 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR1]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8

 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK7-NEXT:    store i32 0, i32* [[R3]], align 4
 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
-// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK7-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
 // CHECK7-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK7:       cond.true:
 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK7-NEXT:    br label [[COND_END:%.*]]
 // CHECK7:       cond.false:

 // CHECK7-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK7-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK7:       omp.inner.for.end:
 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK7:       omp.loop.exit:
 // CHECK7-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP24]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP24]])
 // CHECK7-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 // CHECK7-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
 // CHECK7-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
 // CHECK7-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
 // CHECK7-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK7-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB7:[0-9]+]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
 // CHECK7-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK7-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK7-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK7-NEXT:    ]
 // CHECK7:       .omp.reduction.case1:
 // CHECK7-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK7-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
 // CHECK7-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
 // CHECK7-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
 // CHECK7-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
 // CHECK7-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK7:       cancel.exit:
 // CHECK7-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK7-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
-// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP34]])
+// CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP34]])
 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK7:       .omp.reduction.case2:
 // CHECK7-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
 // CHECK7-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
 // CHECK7-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK7:       .omp.reduction.default:
 // CHECK7-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK7:       omp.precond.end:

 // CHECK8-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK8-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK8-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
-// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
 // CHECK8-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
 // CHECK8-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK8:       omp.inner.for.cond:

 // CHECK8:       .omp.sections.exit:
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK8:       omp.inner.for.inc:
 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK8:       omp.inner.for.end:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK8:       cancel.cont:
-// CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
-// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
+// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK8-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
 // CHECK8-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
 // CHECK8-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
 // CHECK8-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
 // CHECK8:       omp.inner.for.cond6:

 // CHECK8-NEXT:    ]
 // CHECK8:       .omp.sections.case9:
 // CHECK8-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK8-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK8-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
 // CHECK8:       .cancel.exit10:
 // CHECK8-NEXT:    br label [[CANCEL_EXIT19:%.*]]
 // CHECK8:       cancel.exit:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
 // CHECK8:       .cancel.continue11:
 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
 // CHECK8:       .omp.sections.case12:
 // CHECK8-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK8-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK8-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
 // CHECK8:       .cancel.exit13:
 // CHECK8-NEXT:    br label [[CANCEL_EXIT19]]
 // CHECK8:       .cancel.continue14:
 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
 // CHECK8:       .omp.sections.exit15:
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
 // CHECK8:       omp.inner.for.inc16:
 // CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK8-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK8-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND6]]
 // CHECK8:       omp.inner.for.end18:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT20:%.*]]
 // CHECK8:       cancel.cont20:
-// CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
 // CHECK8-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK8-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK8-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK8-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
 // CHECK8-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK8-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK8-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK8-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK8-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
 // CHECK8-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK8:       omp.precond.then:
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK8-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK8-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK8-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK8-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
 // CHECK8-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK8:       cond.true:
 // CHECK8-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
 // CHECK8-NEXT:    br label [[COND_END:%.*]]
 // CHECK8:       cond.false:

 // CHECK8-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK8:       omp_if.then:
 // CHECK8-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
 // CHECK8-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
 // CHECK8-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
 // CHECK8:       .cancel.exit29:
 // CHECK8-NEXT:    br label [[CANCEL_EXIT34:%.*]]
 // CHECK8:       cancel.exit19:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT20]]
 // CHECK8:       .cancel.continue30:
 // CHECK8-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK8:       omp_if.else:
 // CHECK8-NEXT:    br label [[OMP_IF_END]]
 // CHECK8:       omp_if.end:
 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK8:       omp.body.continue:
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
 // CHECK8:       omp.inner.for.inc31:
 // CHECK8-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK8-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
 // CHECK8-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND26]]
 // CHECK8:       omp.inner.for.end33:
 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK8:       omp.loop.exit:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
 // CHECK8-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK8:       cancel.exit34:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT35:%.*]]
 // CHECK8:       omp.precond.end:
 // CHECK8-NEXT:    br label [[CANCEL_CONT35]]
 // CHECK8:       cancel.cont35:
-// CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]])
+// CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK8-NEXT:    [[TMP39:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK8-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
 // CHECK8-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
 // CHECK8-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
 // CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
 // CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*))
 // CHECK8-NEXT:    store i32 0, i32* [[R]], align 4
 // CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])

 // CHECK8-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK8:       omp_if.then:
 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
 // CHECK8-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
 // CHECK8-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
 // CHECK8-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK8:       .cancel.exit:
+// CHECK8-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// CHECK8-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]])
 // CHECK8-NEXT:    br label [[RETURN:%.*]]
 // CHECK8:       .cancel.continue:
 // CHECK8-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK8:       omp_if.else:
 // CHECK8-NEXT:    br label [[OMP_IF_END]]
 // CHECK8:       omp_if.end:
-// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK8-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP7]] to i8
-// CHECK8-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 0
-// CHECK8-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK8-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 0
+// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK8-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
+// CHECK8-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
+// CHECK8-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
+// CHECK8-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK8-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
-// CHECK8-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
-// CHECK8-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP11]])
-// CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
-// CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
+// CHECK8-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
+// CHECK8-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP14]])
+// CHECK8-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
+// CHECK8-NEXT:    br i1 [[TMP16]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
 // CHECK8:       .cancel.exit2:
 // CHECK8-NEXT:    br label [[RETURN]]
 // CHECK8:       .cancel.continue3:
-// CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK8-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK8-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP15]], i64 0
-// CHECK8-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
-// CHECK8-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP16]], i64 0
-// CHECK8-NEXT:    [[TMP17:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
-// CHECK8-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP17]] to i32
-// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP14]]
+// CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK8-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK8-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP18]], i64 0
+// CHECK8-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
+// CHECK8-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP19]], i64 0
+// CHECK8-NEXT:    [[TMP20:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
+// CHECK8-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP20]] to i32
+// CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP17]]
 // CHECK8-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
 // CHECK8-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
 // CHECK8-NEXT:    br label [[RETURN]]
 // CHECK8:       return:
 // CHECK8-NEXT:    ret void
 //
 //
 // CHECK8-LABEL: define {{[^@]+}}@.omp_task_entry.

 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
 // CHECK8-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK8:       omp.inner.for.cond:

 // CHECK8:       .omp.sections.exit:
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK8:       omp.inner.for.inc:
 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK8:       omp.inner.for.end:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK8:       cancel.cont:
 // CHECK8-NEXT:    ret void
 // CHECK8:       cancel.exit:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..3
 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK8-NEXT:  entry:
 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
 // CHECK8-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK8:       omp.inner.for.cond:

 // CHECK8:       .omp.sections.exit:
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK8:       omp.inner.for.inc:
 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK8:       omp.inner.for.end:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK8:       cancel.cont:
 // CHECK8-NEXT:    ret void
 // CHECK8:       cancel.exit:
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..4
 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR1]] {
 // CHECK8-NEXT:  entry:
 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8

 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK8-NEXT:    store i32 0, i32* [[R3]], align 4
 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
-// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK8-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
 // CHECK8-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK8:       cond.true:
 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK8-NEXT:    br label [[COND_END:%.*]]
 // CHECK8:       cond.false:

 // CHECK8-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK8-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK8:       omp.inner.for.end:
 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK8:       omp.loop.exit:
 // CHECK8-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP24]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP24]])
 // CHECK8-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 // CHECK8-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
 // CHECK8-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
 // CHECK8-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
 // CHECK8-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
 // CHECK8-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB7:[0-9]+]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
 // CHECK8-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK8-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK8-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK8-NEXT:    ]
 // CHECK8:       .omp.reduction.case1:
 // CHECK8-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK8-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
 // CHECK8-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
 // CHECK8-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
 // CHECK8-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
 // CHECK8-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK8:       cancel.exit:
 // CHECK8-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK8-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
-// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP34]])
+// CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP34]])
 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK8:       .omp.reduction.case2:
 // CHECK8-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
 // CHECK8-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
 // CHECK8-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK8:       .omp.reduction.default:
 // CHECK8-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK8:       omp.precond.end:

 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
 // CHECK9-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_LASTITER26:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_LOWERBOUND27:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_UPPERBOUND28:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_STRIDE29:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_33:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[I35:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[I36:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
 // CHECK9-NEXT:    [[R:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK9-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK9-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
 // CHECK9-NEXT:    br label [[OMP_PARALLEL:%.*]]
 // CHECK9:       omp_parallel:
 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @main..omp_par to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
 // CHECK9:       omp.par.outlined.exit:
 // CHECK9-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
 // CHECK9:       omp.par.exit.split:
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
 // CHECK9:       omp_section_loop.preheader:
 // CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
 // CHECK9-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
 // CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
 // CHECK9-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
 // CHECK9-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
 // CHECK9:       omp_section_loop.header:
 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
 // CHECK9:       omp_section_loop.cond:
 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
 // CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
 // CHECK9:       omp_section_loop.body:
 // CHECK9-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
 // CHECK9-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
 // CHECK9-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
 // CHECK9-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
 // CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
 // CHECK9-NEXT:    ]
 // CHECK9:       omp_section_loop.inc:
 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
 // CHECK9:       omp_section_loop.exit:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
 // CHECK9:       omp_section_loop.after:
 // CHECK9-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
 // CHECK9:       omp_sections.end:
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER12:%.*]]
-// CHECK9:       omp_section_loop.preheader12:
-// CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND27]], align 4
-// CHECK9-NEXT:    store i32 1, i32* [[P_UPPERBOUND28]], align 4
-// CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE29]], align 4
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM30:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]], i32 34, i32* [[P_LASTITER26]], i32* [[P_LOWERBOUND27]], i32* [[P_UPPERBOUND28]], i32* [[P_STRIDE29]], i32 1, i32 1)
-// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND27]], align 4
-// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND28]], align 4
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
+// CHECK9:       omp_section_loop.preheader13:
+// CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
+// CHECK9-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
+// CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 1)
+// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
+// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
 // CHECK9-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
 // CHECK9-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13:%.*]]
-// CHECK9:       omp_section_loop.header13:
-// CHECK9-NEXT:    [[OMP_SECTION_LOOP_IV19:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER12]] ], [ [[OMP_SECTION_LOOP_NEXT21:%.*]], [[OMP_SECTION_LOOP_INC16:%.*]] ]
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND14:%.*]]
-// CHECK9:       omp_section_loop.cond14:
-// CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP20:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV19]], [[TMP10]]
-// CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP20]], label [[OMP_SECTION_LOOP_BODY15:%.*]], label [[OMP_SECTION_LOOP_EXIT17:%.*]]
-// CHECK9:       omp_section_loop.body15:
-// CHECK9-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV19]], [[TMP7]]
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
+// CHECK9:       omp_section_loop.header14:
+// CHECK9-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
+// CHECK9:       omp_section_loop.cond15:
+// CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
+// CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
+// CHECK9:       omp_section_loop.body16:
+// CHECK9-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
 // CHECK9-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
 // CHECK9-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
-// CHECK9-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC16]] [
-// CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE22:%.*]]
-// CHECK9-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE24:%.*]]
+// CHECK9-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
+// CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
+// CHECK9-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
 // CHECK9-NEXT:    ]
-// CHECK9:       omp_section_loop.inc16:
-// CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT21]] = add nuw i32 [[OMP_SECTION_LOOP_IV19]], 1
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13]]
-// CHECK9:       omp_section_loop.exit17:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]])
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER18:%.*]]
-// CHECK9:       omp_section_loop.after18:
-// CHECK9-NEXT:    br label [[OMP_SECTIONS_END32:%.*]]
-// CHECK9:       omp_sections.end32:
+// CHECK9:       omp_section_loop.inc17:
+// CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
+// CHECK9:       omp_section_loop.exit18:
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
+// CHECK9:       omp_section_loop.after19:
+// CHECK9-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
+// CHECK9:       omp_sections.end33:
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK9-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK9-NEXT:    store i32 [[SUB34]], i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK9-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK9-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK9:       omp.precond.then:
 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK9-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM36:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM36]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
-// CHECK9-NEXT:    [[CMP37:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
-// CHECK9-NEXT:    br i1 [[CMP37]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK9-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
+// CHECK9-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
-// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
-// CHECK9-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK9-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
+// CHECK9-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
-// CHECK9-NEXT:    [[ADD39:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK9-NEXT:    store i32 [[ADD39]], i32* [[I35]], align 4
+// CHECK9-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK9-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
 // CHECK9-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
-// CHECK9-NEXT:    [[TOBOOL40:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
-// CHECK9-NEXT:    br i1 [[TOBOOL40]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
+// CHECK9-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+// CHECK9-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK9:       omp_if.then:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
-// CHECK9-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]], i32 2)
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
+// CHECK9-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
 // CHECK9-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
 // CHECK9-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK9:       .cancel.exit:
 // CHECK9-NEXT:    br label [[CANCEL_EXIT:%.*]]
 // CHECK9:       omp_section_loop.body.case:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM9]], i32 3)
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
 // CHECK9-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
 // CHECK9-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
 // CHECK9:       omp_section_loop.body.case.split:
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
 // CHECK9:       omp_section_loop.body.case.cncl:
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
-// CHECK9:       omp_section_loop.body.case22:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]], i32 3)
+// CHECK9:       omp_section_loop.body.case23:
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
 // CHECK9-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
-// CHECK9-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE22_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE22_CNCL:%.*]]
-// CHECK9:       omp_section_loop.body.case22.split:
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK9:       omp_section_loop.body.case22.cncl:
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK9:       omp_section_loop.body.case24:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM25:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM25]], i32 3)
+// CHECK9-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
+// CHECK9:       omp_section_loop.body.case23.split:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK9:       omp_section_loop.body.case23.cncl:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK9:       omp_section_loop.body.case25:
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
 // CHECK9-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
-// CHECK9-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE24_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE24_CNCL:%.*]]
-// CHECK9:       omp_section_loop.body.case24.split:
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK9:       omp_section_loop.body.case24.cncl:
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
+// CHECK9-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
+// CHECK9:       omp_section_loop.body.case25.split:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK9:       omp_section_loop.body.case25.cncl:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK9:       .cancel.continue:
 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK9:       omp_if.else:
 // CHECK9-NEXT:    br label [[OMP_IF_END]]
 // CHECK9:       omp_if.end:
 // 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:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP35]], 1
-// CHECK9-NEXT:    store i32 [[ADD42]], i32* [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
+// CHECK9-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       cancel.exit:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM43:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM43]])
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
 // CHECK9:       cancel.cont:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
-// CHECK9-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM46]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
+// CHECK9-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK9-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
 // CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
-// CHECK9-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i8* [[TMP36]])
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
+// CHECK9-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
 // CHECK9-NEXT:    store i32 0, i32* [[R]], align 4
 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
 // CHECK9-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK9-NEXT:    ret i32 [[TMP40]]
 //
 //

 // CHECK9:       2:
 // CHECK9-NEXT:    br label [[TMP3:%.*]]
 // CHECK9:       3:
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK9-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
-// CHECK9-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX2]], align 1
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK9-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
+// CHECK9-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
 // CHECK9-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
-// CHECK9-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL4:%.*]]
-// CHECK9:       .cncl4:
+// CHECK9-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
+// CHECK9:       .cncl5:
 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
 // CHECK9:       .cont:
 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
-// CHECK9-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
-// CHECK9-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX5]], align 8
-// CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
-// CHECK9-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX6]], align 1
-// CHECK9-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP12]] to i32
-// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV7]], [[TMP9]]
-// CHECK9-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD]] to i8
-// CHECK9-NEXT:    store i8 [[CONV8]], i8* [[ARRAYIDX6]], align 1
+// CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
+// CHECK9-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
+// CHECK9-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
+// CHECK9-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
+// CHECK9-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP9]]
+// CHECK9-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
+// CHECK9-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
 // CHECK9-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
 // CHECK9:       omp.par.pre_finalize:
 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK9:       13:
 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
 // CHECK9:       .cncl:
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK9:       .split:
 // CHECK9-NEXT:    br label [[TMP3]]
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@.omp_task_entry.
 // CHECK9-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK9-NEXT:  entry:

 // CHECK10-NEXT:  entry:
 // CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
 // CHECK10-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_LASTITER26:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_LOWERBOUND27:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_UPPERBOUND28:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_STRIDE29:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_33:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[I35:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[I36:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
 // CHECK10-NEXT:    [[R:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK10-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK10-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
 // CHECK10-NEXT:    br label [[OMP_PARALLEL:%.*]]
 // CHECK10:       omp_parallel:
 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @main..omp_par to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
 // CHECK10:       omp.par.outlined.exit:
 // CHECK10-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
 // CHECK10:       omp.par.exit.split:
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
 // CHECK10:       omp_section_loop.preheader:
 // CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
 // CHECK10-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
 // CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
 // CHECK10-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
 // CHECK10-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
 // CHECK10:       omp_section_loop.header:
 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
 // CHECK10:       omp_section_loop.cond:
 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
 // CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
 // CHECK10:       omp_section_loop.body:
 // CHECK10-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
 // CHECK10-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
 // CHECK10-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
 // CHECK10-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
 // CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
 // CHECK10-NEXT:    ]
 // CHECK10:       omp_section_loop.inc:
 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
 // CHECK10:       omp_section_loop.exit:
-// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
 // CHECK10:       omp_section_loop.after:
 // CHECK10-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
 // CHECK10:       omp_sections.end:
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER12:%.*]]
-// CHECK10:       omp_section_loop.preheader12:
-// CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND27]], align 4
-// CHECK10-NEXT:    store i32 1, i32* [[P_UPPERBOUND28]], align 4
-// CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE29]], align 4
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM30:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]], i32 34, i32* [[P_LASTITER26]], i32* [[P_LOWERBOUND27]], i32* [[P_UPPERBOUND28]], i32* [[P_STRIDE29]], i32 1, i32 1)
-// CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND27]], align 4
-// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND28]], align 4
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
+// CHECK10:       omp_section_loop.preheader13:
+// CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
+// CHECK10-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
+// CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 1)
+// CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
+// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
 // CHECK10-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
 // CHECK10-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13:%.*]]
-// CHECK10:       omp_section_loop.header13:
-// CHECK10-NEXT:    [[OMP_SECTION_LOOP_IV19:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER12]] ], [ [[OMP_SECTION_LOOP_NEXT21:%.*]], [[OMP_SECTION_LOOP_INC16:%.*]] ]
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND14:%.*]]
-// CHECK10:       omp_section_loop.cond14:
-// CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP20:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV19]], [[TMP10]]
-// CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP20]], label [[OMP_SECTION_LOOP_BODY15:%.*]], label [[OMP_SECTION_LOOP_EXIT17:%.*]]
-// CHECK10:       omp_section_loop.body15:
-// CHECK10-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV19]], [[TMP7]]
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
+// CHECK10:       omp_section_loop.header14:
+// CHECK10-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
+// CHECK10:       omp_section_loop.cond15:
+// CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
+// CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
+// CHECK10:       omp_section_loop.body16:
+// CHECK10-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
 // CHECK10-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
 // CHECK10-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
-// CHECK10-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC16]] [
-// CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE22:%.*]]
-// CHECK10-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE24:%.*]]
+// CHECK10-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
+// CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
+// CHECK10-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
 // CHECK10-NEXT:    ]
-// CHECK10:       omp_section_loop.inc16:
-// CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT21]] = add nuw i32 [[OMP_SECTION_LOOP_IV19]], 1
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13]]
-// CHECK10:       omp_section_loop.exit17:
-// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]])
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER18:%.*]]
-// CHECK10:       omp_section_loop.after18:
-// CHECK10-NEXT:    br label [[OMP_SECTIONS_END32:%.*]]
-// CHECK10:       omp_sections.end32:
+// CHECK10:       omp_section_loop.inc17:
+// CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
+// CHECK10:       omp_section_loop.exit18:
+// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
+// CHECK10:       omp_section_loop.after19:
+// CHECK10-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
+// CHECK10:       omp_sections.end33:
 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK10-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
 // CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK10-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK10-NEXT:    store i32 [[SUB34]], i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK10-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK10-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK10:       omp.precond.then:
 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
-// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK10-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM36:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
-// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM36]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
+// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
-// CHECK10-NEXT:    [[CMP37:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
-// CHECK10-NEXT:    br i1 [[CMP37]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK10-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
+// CHECK10-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK10:       cond.true:
-// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
+// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK10-NEXT:    br label [[COND_END:%.*]]
 // CHECK10:       cond.false:
 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    br label [[COND_END]]
 // CHECK10:       cond.end:
 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK10-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK10:       omp.inner.for.cond:
 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
-// CHECK10-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK10-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
+// CHECK10-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK10:       omp.inner.for.body:
 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
-// CHECK10-NEXT:    [[ADD39:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK10-NEXT:    store i32 [[ADD39]], i32* [[I35]], align 4
+// CHECK10-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK10-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
 // CHECK10-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
-// CHECK10-NEXT:    [[TOBOOL40:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
-// CHECK10-NEXT:    br i1 [[TOBOOL40]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
+// CHECK10-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+// CHECK10-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK10:       omp_if.then:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
-// CHECK10-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]], i32 2)
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
+// CHECK10-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
 // CHECK10-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
 // CHECK10-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK10:       .cancel.exit:
 // CHECK10-NEXT:    br label [[CANCEL_EXIT:%.*]]
 // CHECK10:       omp_section_loop.body.case:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM9]], i32 3)
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
 // CHECK10-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
 // CHECK10-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
 // CHECK10:       omp_section_loop.body.case.split:
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
 // CHECK10:       omp_section_loop.body.case.cncl:
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
-// CHECK10:       omp_section_loop.body.case22:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]], i32 3)
+// CHECK10:       omp_section_loop.body.case23:
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
 // CHECK10-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
-// CHECK10-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE22_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE22_CNCL:%.*]]
-// CHECK10:       omp_section_loop.body.case22.split:
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK10:       omp_section_loop.body.case22.cncl:
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK10:       omp_section_loop.body.case24:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM25:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM25]], i32 3)
+// CHECK10-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
+// CHECK10:       omp_section_loop.body.case23.split:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK10:       omp_section_loop.body.case23.cncl:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK10:       omp_section_loop.body.case25:
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
 // CHECK10-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
-// CHECK10-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE24_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE24_CNCL:%.*]]
-// CHECK10:       omp_section_loop.body.case24.split:
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
-// CHECK10:       omp_section_loop.body.case24.cncl:
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
+// CHECK10-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
+// CHECK10:       omp_section_loop.body.case25.split:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
+// CHECK10:       omp_section_loop.body.case25.cncl:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK10:       .cancel.continue:
 // CHECK10-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK10:       omp_if.else:
 // CHECK10-NEXT:    br label [[OMP_IF_END]]
 // CHECK10:       omp_if.end:
 // 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:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK10-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP35]], 1
-// CHECK10-NEXT:    store i32 [[ADD42]], i32* [[DOTOMP_IV]], align 4
+// CHECK10-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
+// CHECK10-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK10:       omp.inner.for.end:
 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK10:       omp.loop.exit:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
-// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
+// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK10:       cancel.exit:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM43:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
-// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM43]])
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK10:       omp.precond.end:
 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
 // CHECK10:       cancel.cont:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
-// CHECK10-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM46]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
+// CHECK10-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK10-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
 // CHECK10-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
-// CHECK10-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i8* [[TMP36]])
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
+// CHECK10-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
 // CHECK10-NEXT:    store i32 0, i32* [[R]], align 4
 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
 // CHECK10-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK10-NEXT:    ret i32 [[TMP40]]
 //
 //

 // CHECK10:       2:
 // CHECK10-NEXT:    br label [[TMP3:%.*]]
 // CHECK10:       3:
 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
 // CHECK10-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
 // CHECK10-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK10-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
-// CHECK10-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX2]], align 1
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK10-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
+// CHECK10-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
 // CHECK10-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
-// CHECK10-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL4:%.*]]
-// CHECK10:       .cncl4:
+// CHECK10-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
+// CHECK10:       .cncl5:
 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
 // CHECK10:       .cont:
 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK10-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
-// CHECK10-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
-// CHECK10-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX5]], align 8
-// CHECK10-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
-// CHECK10-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX6]], align 1
-// CHECK10-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP12]] to i32
-// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV7]], [[TMP9]]
-// CHECK10-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD]] to i8
-// CHECK10-NEXT:    store i8 [[CONV8]], i8* [[ARRAYIDX6]], align 1
+// CHECK10-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
+// CHECK10-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
+// CHECK10-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
+// CHECK10-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
+// CHECK10-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP12]] to i32
+// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP9]]
+// CHECK10-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
+// CHECK10-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
 // CHECK10-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
 // CHECK10:       omp.par.pre_finalize:
 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK10:       13:
 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK10-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
 // CHECK10-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
 // CHECK10-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
 // CHECK10:       .cncl:
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK10:       .split:
 // CHECK10-NEXT:    br label [[TMP3]]
 //
 //
 // CHECK10-LABEL: define {{[^@]+}}@.omp_task_entry.
 // CHECK10-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK10-NEXT:  entry:

 // CHECK10-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
 // CHECK10-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
 // CHECK10-NEXT:    ret void
 //
-//
 

clang/test/OpenMP/cancellation_point_codegen.cpp

 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK1-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK1-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
 // CHECK1-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:

 // CHECK1:       .omp.sections.exit:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK1:       cancel.cont:
-// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_3]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_4]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_5]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_6]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_6]], i32* [[DOTOMP_SECTIONS_LB_3]], i32* [[DOTOMP_SECTIONS_UB_4]], i32* [[DOTOMP_SECTIONS_ST_5]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_6]], i32* [[DOTOMP_SECTIONS_LB_3]], i32* [[DOTOMP_SECTIONS_UB_4]], i32* [[DOTOMP_SECTIONS_ST_5]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_4]], align 4
 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp slt i32 [[TMP13]], 1
 // CHECK1-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP13]], i32 1
 // CHECK1-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_SECTIONS_UB_4]], align 4
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_3]], align 4
 // CHECK1-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_SECTIONS_IV_7]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
 // CHECK1:       omp.inner.for.cond8:

 // CHECK1-NEXT:    ]
 // CHECK1:       .omp.sections.case11:
 // CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancellationpoint(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK1-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK1-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT12:%.*]], label [[DOTCANCEL_CONTINUE13:%.*]]
 // CHECK1:       .cancel.exit12:
 // CHECK1-NEXT:    br label [[CANCEL_EXIT23:%.*]]
 // CHECK1:       cancel.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
 // CHECK1:       .cancel.continue13:
 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT19]]
 // CHECK1:       .omp.sections.case14:
 // CHECK1-NEXT:    [[TMP22:%.*]] = call i32 @__kmpc_cancellationpoint(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK1-NEXT:    br i1 [[TMP23]], label [[DOTCANCEL_EXIT15:%.*]], label [[DOTCANCEL_CONTINUE16:%.*]]
 // CHECK1:       .cancel.exit15:

 // CHECK1:       .omp.sections.exit19:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC20:%.*]]
 // CHECK1:       omp.inner.for.inc20:
 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_7]], align 4
 // CHECK1-NEXT:    [[INC21:%.*]] = add nsw i32 [[TMP26]], 1
 // CHECK1-NEXT:    store i32 [[INC21]], i32* [[DOTOMP_SECTIONS_IV_7]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND8]]
 // CHECK1:       omp.inner.for.end22:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT24:%.*]]
 // CHECK1:       cancel.cont24:
-// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK1-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK1-NEXT:    store i32 [[SUB26]], i32* [[DOTCAPTURE_EXPR_25]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[CMP27:%.*]] = icmp slt i32 0, [[TMP29]]
 // CHECK1-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
 // CHECK1-NEXT:    store i32 [[TMP30]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
 // CHECK1-NEXT:    [[CMP29:%.*]] = icmp sgt i32 [[TMP31]], [[TMP32]]
 // CHECK1-NEXT:    br i1 [[CMP29]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:

 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I28]], align 4
 // CHECK1-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_cancellationpoint(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
 // CHECK1-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
 // CHECK1-NEXT:    br i1 [[TMP40]], label [[DOTCANCEL_EXIT33:%.*]], label [[DOTCANCEL_CONTINUE34:%.*]]
 // CHECK1:       .cancel.exit33:
 // CHECK1-NEXT:    br label [[CANCEL_EXIT40:%.*]]
 // CHECK1:       cancel.exit23:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT24]]
 // CHECK1:       .cancel.continue34:
 // CHECK1-NEXT:    [[TMP41:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
 // CHECK1-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
 // CHECK1-NEXT:    br i1 [[TMP42]], label [[DOTCANCEL_EXIT35:%.*]], label [[DOTCANCEL_CONTINUE36:%.*]]
 // CHECK1:       .cancel.exit35:
 // CHECK1-NEXT:    br label [[CANCEL_EXIT40]]
 // CHECK1:       .cancel.continue36:
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
 // CHECK1:       omp.inner.for.inc37:
 // CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP43]], 1
 // CHECK1-NEXT:    store i32 [[ADD38]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND30]]
 // CHECK1:       omp.inner.for.end39:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       cancel.exit40:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT41:%.*]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    br label [[CANCEL_CONT41]]
 // CHECK1:       cancel.cont41:
-// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK1-NEXT:    [[TMP44:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK1-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP44]] to %struct.kmp_task_t_with_privates*
 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP45]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP47:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP44]])
 // CHECK1-NEXT:    [[TMP48:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates.1*)* @.omp_task_entry..3 to i32 (i32, i8*)*))
 // CHECK1-NEXT:    [[TMP49:%.*]] = bitcast i8* [[TMP48]] to %struct.kmp_task_t_with_privates.1*
 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES_1:%.*]], %struct.kmp_task_t_with_privates.1* [[TMP49]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP51:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP48]])

 // CHECK1-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_cancellationpoint(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 1)
 // CHECK1-NEXT:    [[TMP5:%.*]] = icmp ne i32 [[TMP4]], 0
 // CHECK1-NEXT:    br i1 [[TMP5]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK1:       .cancel.exit:
+// CHECK1-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]])
 // CHECK1-NEXT:    br label [[RETURN:%.*]]
 // CHECK1:       .cancel.continue:
-// CHECK1-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 1)
-// CHECK1-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
-// CHECK1-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT1:%.*]], label [[DOTCANCEL_CONTINUE2:%.*]]
+// CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 1)
+// CHECK1-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
+// CHECK1-NEXT:    br i1 [[TMP8]], label [[DOTCANCEL_EXIT1:%.*]], label [[DOTCANCEL_CONTINUE2:%.*]]
 // CHECK1:       .cancel.exit1:
+// CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]])
 // CHECK1-NEXT:    br label [[RETURN]]
 // CHECK1:       .cancel.continue2:
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP8]] to i8
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP9]], i64 0
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP10]], i64 0
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
+// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK1-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
 // CHECK1-NEXT:    br label [[RETURN]]
 // CHECK1:       return:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {

 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:

 // CHECK1:       .omp.sections.exit:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK1:       cancel.cont:
 // CHECK1-NEXT:    ret void
 // CHECK1:       cancel.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
 // CHECK1-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:

 // CHECK1:       .omp.sections.exit:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP15]], 1
 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK1:       cancel.cont:
 // CHECK1-NEXT:    ret void
 // CHECK1:       cancel.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8

 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:

 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], 1
 // CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP25]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP25]])
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       cancel.exit:
 // 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* @[[GLOB4]], i32 [[TMP27]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP27]])
 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
 // CHECK1:       cancel.cont:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@main

 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
 // CHECK2-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:

 // CHECK2:       .omp.sections.exit:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK2:       cancel.cont:
-// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_3]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_4]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_5]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_6]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_6]], i32* [[DOTOMP_SECTIONS_LB_3]], i32* [[DOTOMP_SECTIONS_UB_4]], i32* [[DOTOMP_SECTIONS_ST_5]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_6]], i32* [[DOTOMP_SECTIONS_LB_3]], i32* [[DOTOMP_SECTIONS_UB_4]], i32* [[DOTOMP_SECTIONS_ST_5]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_4]], align 4
 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp slt i32 [[TMP13]], 1
 // CHECK2-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP13]], i32 1
 // CHECK2-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_SECTIONS_UB_4]], align 4
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_3]], align 4
 // CHECK2-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_SECTIONS_IV_7]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
 // CHECK2:       omp.inner.for.cond8:

 // CHECK2-NEXT:    ]
 // CHECK2:       .omp.sections.case11:
 // CHECK2-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancellationpoint(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK2-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK2-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT12:%.*]], label [[DOTCANCEL_CONTINUE13:%.*]]
 // CHECK2:       .cancel.exit12:
 // CHECK2-NEXT:    br label [[CANCEL_EXIT23:%.*]]
 // CHECK2:       cancel.exit:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
 // CHECK2:       .cancel.continue13:
 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT19]]
 // CHECK2:       .omp.sections.case14:
 // CHECK2-NEXT:    [[TMP22:%.*]] = call i32 @__kmpc_cancellationpoint(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
 // CHECK2-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK2-NEXT:    br i1 [[TMP23]], label [[DOTCANCEL_EXIT15:%.*]], label [[DOTCANCEL_CONTINUE16:%.*]]
 // CHECK2:       .cancel.exit15:

 // CHECK2:       .omp.sections.exit19:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC20:%.*]]
 // CHECK2:       omp.inner.for.inc20:
 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_7]], align 4
 // CHECK2-NEXT:    [[INC21:%.*]] = add nsw i32 [[TMP26]], 1
 // CHECK2-NEXT:    store i32 [[INC21]], i32* [[DOTOMP_SECTIONS_IV_7]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND8]]
 // CHECK2:       omp.inner.for.end22:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT24:%.*]]
 // CHECK2:       cancel.cont24:
-// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK2-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK2-NEXT:    store i32 [[SUB26]], i32* [[DOTCAPTURE_EXPR_25]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK2-NEXT:    [[CMP27:%.*]] = icmp slt i32 0, [[TMP29]]
 // CHECK2-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK2:       omp.precond.then:
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
 // CHECK2-NEXT:    store i32 [[TMP30]], i32* [[DOTOMP_UB]], 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:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
 // CHECK2-NEXT:    [[CMP29:%.*]] = icmp sgt i32 [[TMP31]], [[TMP32]]
 // CHECK2-NEXT:    br i1 [[CMP29]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:

 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I28]], align 4
 // CHECK2-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_cancellationpoint(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
 // CHECK2-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
 // CHECK2-NEXT:    br i1 [[TMP40]], label [[DOTCANCEL_EXIT33:%.*]], label [[DOTCANCEL_CONTINUE34:%.*]]
 // CHECK2:       .cancel.exit33:
 // CHECK2-NEXT:    br label [[CANCEL_EXIT40:%.*]]
 // CHECK2:       cancel.exit23:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT24]]
 // CHECK2:       .cancel.continue34:
 // CHECK2-NEXT:    [[TMP41:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
 // CHECK2-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
 // CHECK2-NEXT:    br i1 [[TMP42]], label [[DOTCANCEL_EXIT35:%.*]], label [[DOTCANCEL_CONTINUE36:%.*]]
 // CHECK2:       .cancel.exit35:
 // CHECK2-NEXT:    br label [[CANCEL_EXIT40]]
 // CHECK2:       .cancel.continue36:
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
 // CHECK2:       omp.inner.for.inc37:
 // CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP43]], 1
 // CHECK2-NEXT:    store i32 [[ADD38]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND30]]
 // CHECK2:       omp.inner.for.end39:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       cancel.exit40:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT41:%.*]]
 // CHECK2:       omp.precond.end:
 // CHECK2-NEXT:    br label [[CANCEL_CONT41]]
 // CHECK2:       cancel.cont41:
-// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK2-NEXT:    [[TMP44:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
 // CHECK2-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP44]] to %struct.kmp_task_t_with_privates*
 // CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP45]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP47:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP44]])
 // CHECK2-NEXT:    [[TMP48:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates.1*)* @.omp_task_entry..3 to i32 (i32, i8*)*))
 // CHECK2-NEXT:    [[TMP49:%.*]] = bitcast i8* [[TMP48]] to %struct.kmp_task_t_with_privates.1*
 // CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES_1:%.*]], %struct.kmp_task_t_with_privates.1* [[TMP49]], i32 0, i32 0
 // CHECK2-NEXT:    [[TMP51:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP48]])

 // CHECK2-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_cancellationpoint(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 1)
 // CHECK2-NEXT:    [[TMP5:%.*]] = icmp ne i32 [[TMP4]], 0
 // CHECK2-NEXT:    br i1 [[TMP5]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK2:       .cancel.exit:
+// CHECK2-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]])
 // CHECK2-NEXT:    br label [[RETURN:%.*]]
 // CHECK2:       .cancel.continue:
-// CHECK2-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 1)
-// CHECK2-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
-// CHECK2-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT1:%.*]], label [[DOTCANCEL_CONTINUE2:%.*]]
+// CHECK2-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 1)
+// CHECK2-NEXT:    [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
+// CHECK2-NEXT:    br i1 [[TMP8]], label [[DOTCANCEL_EXIT1:%.*]], label [[DOTCANCEL_CONTINUE2:%.*]]
 // CHECK2:       .cancel.exit1:
+// CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]])
 // CHECK2-NEXT:    br label [[RETURN]]
 // CHECK2:       .cancel.continue2:
-// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP8]] to i8
-// CHECK2-NEXT:    [[TMP9:%.*]] = load i8**, i8*** [[TMP0]], align 8
-// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP9]], i64 0
-// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
-// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP10]], i64 0
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
+// CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
+// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
+// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK2-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
 // CHECK2-NEXT:    br label [[RETURN]]
 // CHECK2:       return:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {

 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
 // CHECK2-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:

 // CHECK2:       .omp.sections.exit:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK2:       cancel.cont:
 // CHECK2-NEXT:    ret void
 // CHECK2:       cancel.exit:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
 // CHECK2-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:

 // CHECK2:       .omp.sections.exit:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP15]], 1
 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK2:       cancel.cont:
 // CHECK2-NEXT:    ret void
 // CHECK2:       cancel.exit:
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP1]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8

 // CHECK2:       omp.precond.then:
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], 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:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK2-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:

 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP23]], 1
 // CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP25]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP25]])
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       cancel.exit:
 // 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* @[[GLOB4]], i32 [[TMP27]])
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP27]])
 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK2:       omp.precond.end:
 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
 // CHECK2:       cancel.cont:
 // CHECK2-NEXT:    ret void
 //
-//
 

clang/test/OpenMP/distribute_parallel_for_codegen.cpp

 //
 //
 // CHECK12-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
 // CHECK12-SAME: () #[[ATTR4:[0-9]+]] {
 // CHECK12-NEXT:  entry:
 // CHECK12-NEXT:    call void @__tgt_register_requires(i64 1)
 // CHECK12-NEXT:    ret void
 //
-//
+//
 

clang/test/OpenMP/parallel_master_taskloop_codegen.cpp

 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_parallel_master_taskloop_codegen.cpp
 // CHECK2-SAME: () #[[ATTR7]] section "__TEXT,__StaticInit,regular,pure_instructions" {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    call void @__cxx_global_var_init()
 // CHECK2-NEXT:    ret void
 //
-//
 

clang/test/OpenMP/target_parallel_codegen.cpp

 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK1-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK1-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK1-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK1:       .cancel.exit:
+// CHECK1-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK1-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK1:       .cancel.continue:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK1-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
 // CHECK1-NEXT:  entry:

 // CHECK2-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK2-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK2-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK2-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK2:       .cancel.exit:
+// CHECK2-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK2-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK2:       .cancel.continue:
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK2-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
 // CHECK2-NEXT:  entry:

 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK3-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK3-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK3-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK3:       .cancel.exit:
+// CHECK3-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK3-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK3:       .cancel.continue:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK3-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
 // CHECK3-NEXT:  entry:

 // CHECK4-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK4-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK4-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK4-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK4-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK4:       .cancel.exit:
+// CHECK4-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK4-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK4:       .cancel.continue:
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK4-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
 // CHECK4-NEXT:  entry:

 // CHECK9-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK9-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK9-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK9-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK9-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK9:       .cancel.exit:
+// CHECK9-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK9-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK9:       .cancel.continue:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK9-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:

 // CHECK10-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK10-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK10-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK10-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK10-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK10:       .cancel.exit:
+// CHECK10-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK10-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK10:       .cancel.continue:
 // CHECK10-NEXT:    ret void
 //
 //
 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK10-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK10-NEXT:  entry:

 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK11-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK11-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK11-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK11-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK11:       .cancel.exit:
+// CHECK11-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK11-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK11:       .cancel.continue:
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK11-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:

 // CHECK12-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK12-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK12-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK12-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK12-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK12:       .cancel.exit:
+// CHECK12-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK12-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK12:       .cancel.continue:
 // CHECK12-NEXT:    ret void
 //
 //
 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK12-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK12-NEXT:  entry:

 // CHECK17-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK17-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK17-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK17-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK17-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK17:       .cancel.exit:
+// CHECK17-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK17-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK17:       .cancel.continue:
 // CHECK17-NEXT:    ret void
 //
 //
 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK17-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
 // CHECK17-NEXT:  entry:

 // CHECK18-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK18-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK18-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK18-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK18-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK18:       .cancel.exit:
+// CHECK18-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK18-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK18:       .cancel.continue:
 // CHECK18-NEXT:    ret void
 //
 //
 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK18-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
 // CHECK18-NEXT:  entry:

 // CHECK19-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK19-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK19-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK19-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK19-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK19:       .cancel.exit:
+// CHECK19-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK19-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK19:       .cancel.continue:
 // CHECK19-NEXT:    ret void
 //
 //
 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK19-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
 // CHECK19-NEXT:  entry:

 // CHECK20-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK20-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK20-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK20-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK20-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK20:       .cancel.exit:
+// CHECK20-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK20-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK20:       .cancel.continue:
 // CHECK20-NEXT:    ret void
 //
 //
 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK20-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
 // CHECK20-NEXT:  entry:

 // CHECK25-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK25-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK25-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK25-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK25-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK25:       .cancel.exit:
+// CHECK25-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK25-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK25:       .cancel.continue:
 // CHECK25-NEXT:    ret void
 //
 //
 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK25-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK25-NEXT:  entry:

 // CHECK26-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK26-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK26-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK26-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK26-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK26:       .cancel.exit:
+// CHECK26-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK26-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK26:       .cancel.continue:
 // CHECK26-NEXT:    ret void
 //
 //
 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK26-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK26-NEXT:  entry:

 // CHECK27-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK27-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK27-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK27-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK27-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK27:       .cancel.exit:
+// CHECK27-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK27-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK27:       .cancel.continue:
 // CHECK27-NEXT:    ret void
 //
 //
 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK27-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK27-NEXT:  entry:

 // CHECK28-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
 // CHECK28-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
 // CHECK28-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
 // CHECK28-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
 // CHECK28-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK28:       .cancel.exit:
+// CHECK28-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
 // CHECK28-NEXT:    br label [[DOTCANCEL_CONTINUE]]
 // CHECK28:       .cancel.continue:
 // CHECK28-NEXT:    ret void
 //
 //
 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
 // CHECK28-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
 // CHECK28-NEXT:  entry:

 // CHECK28-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
 // CHECK28-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
 // CHECK28-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
 // CHECK28-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
 // CHECK28-NEXT:    ret void
 //
-//
+//
 

clang/test/OpenMP/target_parallel_for_codegen.cpp

 // CHECK28-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8
 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK28:       omp.inner.for.end:
 // CHECK28-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK28:       omp.loop.exit:
 // CHECK28-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
 // CHECK28-NEXT:    ret void
 //
-//
+//
 

clang/test/OpenMP/target_teams_distribute_parallel_for_codegen.cpp

 // CHECK13:       omp.loop.exit:
 // CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
 // CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
 // CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK13:       omp.precond.end:
 // CHECK13-NEXT:    ret void
 //
-//
 

clang/test/OpenMP/teams_distribute_parallel_for_codegen.cpp

 //
 //
 // CHECK28-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
 // CHECK28-SAME: () #[[ATTR5:[0-9]+]] {
 // CHECK28-NEXT:  entry:
 // CHECK28-NEXT:    call void @__tgt_register_requires(i64 1)
 // CHECK28-NEXT:    ret void
 //
-//
 

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

public:
 
   // Get the type corresponding to __kmpc_impl_lanemask_t from the deviceRTL
   Type *getLanemaskType();
 
   /// Generate control flow and cleanup for cancellation.
   ///
   /// \param CancelFlag Flag indicating if the cancellation is performed.
   /// \param CanceledDirective The kind of directive that is cancled.
+  /// \param ExitCB Extra code to be generated in the exit block.
   void emitCancelationCheckImpl(Value *CancelFlag,
-                                omp::Directive CanceledDirective);
+                                omp::Directive CanceledDirective,
+                                FinalizeCallbackTy ExitCB = {});
 
   /// Generate a barrier runtime call.
   ///
   /// \param Loc The location at which the request originated and is fulfilled.
   /// \param DK The directive which caused the barrier
   /// \param ForceSimpleCall Flag to force a simple (=non-cancellation) barrier.
   /// \param CheckCancelFlag Flag to indicate a cancel barrier return value
   ///                        should be checked and acted upon.

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

default:
     llvm_unreachable("Unknown cancel kind!");
   }
 
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc);
   Value *Ident = getOrCreateIdent(SrcLocStr);
   Value *Args[] = {Ident, getOrCreateThreadID(Ident), CancelKind};
   Value *Result = Builder.CreateCall(
       getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_cancel), Args);
+  auto ExitCB = [this, CanceledDirective, Loc](InsertPointTy IP) {
+    if (CanceledDirective == OMPD_parallel) {
+      IRBuilder<>::InsertPointGuard IPG(Builder);
+      Builder.restoreIP(IP);
+      createBarrier(LocationDescription(Builder.saveIP(), Loc.DL),
+                    omp::Directive::OMPD_unknown, /* ForceSimpleCall */ false,
+                    /* CheckCancelFlag */ false);
+    }
+  };
 
   // The actual cancel logic is shared with others, e.g., cancel_barriers.
-  emitCancelationCheckImpl(Result, CanceledDirective);
+  emitCancelationCheckImpl(Result, CanceledDirective, ExitCB);
 
   // Update the insertion point and remove the terminator we introduced.
   Builder.SetInsertPoint(UI->getParent());
   UI->eraseFromParent();
 
   return Builder.saveIP();
 }
 
-void OpenMPIRBuilder::emitCancelationCheckImpl(
-    Value *CancelFlag, omp::Directive CanceledDirective) {
+void OpenMPIRBuilder::emitCancelationCheckImpl(Value *CancelFlag,
+                                               omp::Directive CanceledDirective,
+                                               FinalizeCallbackTy ExitCB) {
   assert(isLastFinalizationInfoCancellable(CanceledDirective) &&
          "Unexpected cancellation!");
 
   // For a cancel barrier we create two new blocks.
   BasicBlock *BB = Builder.GetInsertBlock();
   BasicBlock *NonCancellationBlock;
   if (Builder.GetInsertPoint() == BB->end()) {
     // TODO: This branch will not be needed once we moved to the
void OpenMPIRBuilder::emitCancelationCheckImpl(Value *CancelFlag,
   // Jump to them based on the return value.
   Value *Cmp = Builder.CreateIsNull(CancelFlag);
   Builder.CreateCondBr(Cmp, NonCancellationBlock, CancellationBlock,
                        /* TODO weight */ nullptr, nullptr);
 
   // From the cancellation block we finalize all variables and go to the
   // post finalization block that is known to the FiniCB callback.
   Builder.SetInsertPoint(CancellationBlock);
+  if (ExitCB)
+    ExitCB(Builder.saveIP());
   auto &FI = FinalizationStack.back();
   FI.FiniCB(Builder.saveIP());
 
   // The continuation block is where code generation continues.
   Builder.SetInsertPoint(NonCancellationBlock, NonCancellationBlock->begin());
 }
 
 IRBuilder<>::InsertPoint OpenMPIRBuilder::createParallel(

llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp

TEST_F(OpenMPIRBuilderTest, CreateCancel) {
   OMPBuilder.pushFinalizationCB({FiniCB, OMPD_parallel, true});
 
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP()});
   auto NewIP = OMPBuilder.createCancel(Loc, nullptr, OMPD_parallel);
   Builder.restoreIP(NewIP);
   EXPECT_FALSE(M->global_empty());
-  EXPECT_EQ(M->size(), 3U);
+  EXPECT_EQ(M->size(), 4U);
   EXPECT_EQ(F->size(), 4U);
   EXPECT_EQ(BB->size(), 4U);
 
   CallInst *GTID = dyn_cast<CallInst>(&BB->front());
   EXPECT_NE(GTID, nullptr);
   EXPECT_EQ(GTID->getNumArgOperands(), 1U);
   EXPECT_EQ(GTID->getCalledFunction()->getName(), "__kmpc_global_thread_num");
   EXPECT_FALSE(GTID->getCalledFunction()->doesNotAccessMemory());
   EXPECT_FALSE(GTID->getCalledFunction()->doesNotFreeMemory());
 
   CallInst *Cancel = dyn_cast<CallInst>(GTID->getNextNode());
   EXPECT_NE(Cancel, nullptr);
   EXPECT_EQ(Cancel->getNumArgOperands(), 3U);
   EXPECT_EQ(Cancel->getCalledFunction()->getName(), "__kmpc_cancel");
   EXPECT_FALSE(Cancel->getCalledFunction()->doesNotAccessMemory());
   EXPECT_FALSE(Cancel->getCalledFunction()->doesNotFreeMemory());
   EXPECT_EQ(Cancel->getNumUses(), 1U);
   Instruction *CancelBBTI = Cancel->getParent()->getTerminator();
   EXPECT_EQ(CancelBBTI->getNumSuccessors(), 2U);
   EXPECT_EQ(CancelBBTI->getSuccessor(0), NewIP.getBlock());
-  EXPECT_EQ(CancelBBTI->getSuccessor(1)->size(), 1U);
+  EXPECT_EQ(CancelBBTI->getSuccessor(1)->size(), 3U);
+  CallInst *GTID1 = dyn_cast<CallInst>(&CancelBBTI->getSuccessor(1)->front());
+  EXPECT_NE(GTID1, nullptr);
+  EXPECT_EQ(GTID1->getNumArgOperands(), 1U);
+  EXPECT_EQ(GTID1->getCalledFunction()->getName(), "__kmpc_global_thread_num");
+  EXPECT_FALSE(GTID1->getCalledFunction()->doesNotAccessMemory());
+  EXPECT_FALSE(GTID1->getCalledFunction()->doesNotFreeMemory());
+  CallInst *Barrier = dyn_cast<CallInst>(GTID1->getNextNode());
+  EXPECT_NE(Barrier, nullptr);
+  EXPECT_EQ(Barrier->getNumArgOperands(), 2U);
+  EXPECT_EQ(Barrier->getCalledFunction()->getName(), "__kmpc_cancel_barrier");
+  EXPECT_FALSE(Barrier->getCalledFunction()->doesNotAccessMemory());
+  EXPECT_FALSE(Barrier->getCalledFunction()->doesNotFreeMemory());
+  EXPECT_EQ(Barrier->getNumUses(), 0U);
   EXPECT_EQ(CancelBBTI->getSuccessor(1)->getTerminator()->getNumSuccessors(),
             1U);
   EXPECT_EQ(CancelBBTI->getSuccessor(1)->getTerminator()->getSuccessor(0),
             CBB);
 
   EXPECT_EQ(cast<CallInst>(Cancel)->getArgOperand(1), GTID);
 
   OMPBuilder.popFinalizationCB();
TEST_F(OpenMPIRBuilderTest, CreateCancelIfCond) {
   OMPBuilder.pushFinalizationCB({FiniCB, OMPD_parallel, true});
 
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP()});
   auto NewIP = OMPBuilder.createCancel(Loc, Builder.getTrue(), OMPD_parallel);
   Builder.restoreIP(NewIP);
   EXPECT_FALSE(M->global_empty());
-  EXPECT_EQ(M->size(), 3U);
+  EXPECT_EQ(M->size(), 4U);
   EXPECT_EQ(F->size(), 7U);
   EXPECT_EQ(BB->size(), 1U);
   ASSERT_TRUE(isa<BranchInst>(BB->getTerminator()));
   ASSERT_EQ(BB->getTerminator()->getNumSuccessors(), 2U);
   BB = BB->getTerminator()->getSuccessor(0);
   EXPECT_EQ(BB->size(), 4U);
 
 
TEST_F(OpenMPIRBuilderTest, CreateCancelIfCond) {
   EXPECT_EQ(Cancel->getCalledFunction()->getName(), "__kmpc_cancel");
   EXPECT_FALSE(Cancel->getCalledFunction()->doesNotAccessMemory());
   EXPECT_FALSE(Cancel->getCalledFunction()->doesNotFreeMemory());
   EXPECT_EQ(Cancel->getNumUses(), 1U);
   Instruction *CancelBBTI = Cancel->getParent()->getTerminator();
   EXPECT_EQ(CancelBBTI->getNumSuccessors(), 2U);
   EXPECT_EQ(CancelBBTI->getSuccessor(0)->size(), 1U);
   EXPECT_EQ(CancelBBTI->getSuccessor(0)->getUniqueSuccessor(), NewIP.getBlock());
-  EXPECT_EQ(CancelBBTI->getSuccessor(1)->size(), 1U);
+  EXPECT_EQ(CancelBBTI->getSuccessor(1)->size(), 3U);
+  CallInst *GTID1 = dyn_cast<CallInst>(&CancelBBTI->getSuccessor(1)->front());
+  EXPECT_NE(GTID1, nullptr);
+  EXPECT_EQ(GTID1->getNumArgOperands(), 1U);
+  EXPECT_EQ(GTID1->getCalledFunction()->getName(), "__kmpc_global_thread_num");
+  EXPECT_FALSE(GTID1->getCalledFunction()->doesNotAccessMemory());
+  EXPECT_FALSE(GTID1->getCalledFunction()->doesNotFreeMemory());
+  CallInst *Barrier = dyn_cast<CallInst>(GTID1->getNextNode());
+  EXPECT_NE(Barrier, nullptr);
+  EXPECT_EQ(Barrier->getNumArgOperands(), 2U);
+  EXPECT_EQ(Barrier->getCalledFunction()->getName(), "__kmpc_cancel_barrier");
+  EXPECT_FALSE(Barrier->getCalledFunction()->doesNotAccessMemory());
+  EXPECT_FALSE(Barrier->getCalledFunction()->doesNotFreeMemory());
+  EXPECT_EQ(Barrier->getNumUses(), 0U);
   EXPECT_EQ(CancelBBTI->getSuccessor(1)->getTerminator()->getNumSuccessors(),
             1U);
   EXPECT_EQ(CancelBBTI->getSuccessor(1)->getTerminator()->getSuccessor(0),
             CBB);
 
   EXPECT_EQ(cast<CallInst>(Cancel)->getArgOperand(1), GTID);
 
   OMPBuilder.popFinalizationCB();