// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
// Check code generation
// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1

// Check same results after serialization round-trip
// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK2
// expected-no-diagnostics

#ifndef HEADER
#define HEADER

// placeholder for loop body code.
extern "C" void body(...) {}


struct S {
  int i;
  S() {
#pragma omp tile sizes(5)
    for (i = 7; i < 17; i += 3)
      body(i);
  }
} s;

extern "C" void foo1(int start, int end, int step) {
  int i;
#pragma omp tile sizes(5)
  for (i = start; i < end; i += step)
    body(i);
}

extern "C" void foo2(int start, int end, int step) {
#pragma omp tile sizes(5,5)
  for (int i = 7; i < 17; i+=3)
    for (int j = 7; j < 17; j+=3)
      body(i,j);
}

extern "C" void foo3() {
#pragma omp for
#pragma omp tile sizes(5,5)
    for (int i = 7; i < 17; i += 3)
      for (int j = 7; j < 17; j += 3)
        body(i, j);
}

extern "C" void foo4() {
#pragma omp for collapse(2)
  for (int k = 7; k < 17; k += 3)
#pragma omp tile sizes(5,5)
  for (int i = 7; i < 17; i += 3)
    for (int j = 7; j < 17; j += 3)
      body(i, j);
}


extern "C" void foo5() {
#pragma omp for collapse(3)
#pragma omp tile sizes(5)
  for (int i = 7; i < 17; i += 3)
    for (int j = 7; j < 17; j += 3)
      body(i, j);
}


extern "C" void foo6() {
#pragma omp parallel for
#pragma omp tile sizes(5)
  for (int i = 7; i < 17; i += 3)
    body(i);
}


template<typename T, T Step, T Tile>
void foo7(T start, T end) {
#pragma omp tile sizes(Tile)
  for (T i = start; i < end; i += Step)
    body(i);
}

extern "C" void tfoo7() {
  foo7<int,3,5>(0, 42);
}

#endif /* HEADER */
// CHECK1-LABEL: define {{[^@]+}}@body
// CHECK1-SAME: (...) #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@__cxx_global_var_init
// CHECK1-SAME: () #[[ATTR1:[0-9]+]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @_ZN1SC1Ev(%struct.S* nonnull align 4 dereferenceable(4) @s)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SC1Ev
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR2:[0-9]+]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SC2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SC2Ev
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR2]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[I2:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    store i32* [[I2]], i32** [[I]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
// CHECK1:       for.cond:
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], 4
// CHECK1-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END11:%.*]]
// CHECK1:       for.body:
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK1:       for.cond3:
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 5
// CHECK1-NEXT:    [[CMP4:%.*]] = icmp slt i32 4, [[ADD]]
// CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP4]], 5
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[ADD5]], [[COND_FALSE]] ]
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP2]], [[COND]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[FOR_BODY7:%.*]], label [[FOR_END:%.*]]
// CHECK1:       for.body7:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP5]], 3
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 7, [[MUL]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[I]], align 8
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[TMP6]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[I]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK1-NEXT:    call void (...) @body(i32 [[TMP8]])
// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
// CHECK1:       for.inc:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK1:       for.end:
// CHECK1-NEXT:    br label [[FOR_INC9:%.*]]
// CHECK1:       for.inc9:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK1-NEXT:    store i32 [[ADD10]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK1:       for.end11:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@foo1
// CHECK1-SAME: (i32 [[START:%.*]], i32 [[END:%.*]], i32 [[STEP:%.*]]) #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[START_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[END_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[STEP_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32 [[START]], i32* [[START_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[END]], i32* [[END_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[STEP]], i32* [[STEP_ADDR]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[START_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[END_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[STEP_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
// CHECK1-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], [[TMP5]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], [[TMP6]]
// CHECK1-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
// CHECK1-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
// CHECK1:       for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[ADD6:%.*]] = add i32 [[TMP8]], 1
// CHECK1-NEXT:    [[CMP:%.*]] = icmp ult i32 [[TMP7]], [[ADD6]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END18:%.*]]
// CHECK1:       for.body:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 [[TMP9]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND7:%.*]]
// CHECK1:       for.cond7:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add i32 [[TMP11]], 1
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP12]], 5
// CHECK1-NEXT:    [[CMP10:%.*]] = icmp ult i32 [[ADD8]], [[ADD9]]
// CHECK1-NEXT:    br i1 [[CMP10]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[ADD11:%.*]] = add i32 [[TMP13]], 1
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP14]], 5
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[ADD11]], [[COND_TRUE]] ], [ [[ADD12]], [[COND_FALSE]] ]
// CHECK1-NEXT:    [[CMP13:%.*]] = icmp ult i32 [[TMP10]], [[COND]]
// CHECK1-NEXT:    br i1 [[CMP13]], label [[FOR_BODY14:%.*]], label [[FOR_END:%.*]]
// CHECK1:       for.body14:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], [[TMP17]]
// CHECK1-NEXT:    [[ADD15:%.*]] = add i32 [[TMP15]], [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD15]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    call void (...) @body(i32 [[TMP18]])
// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
// CHECK1:       for.inc:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP19]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND7]], !llvm.loop [[LOOP5:![0-9]+]]
// CHECK1:       for.end:
// CHECK1-NEXT:    br label [[FOR_INC16:%.*]]
// CHECK1:       for.inc16:
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 5
// CHECK1-NEXT:    store i32 [[ADD17]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
// CHECK1:       for.end18:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@foo2
// CHECK1-SAME: (i32 [[START:%.*]], i32 [[END:%.*]], i32 [[STEP:%.*]]) #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[START_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[END_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[STEP_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_1_IV_J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_1_IV_J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32 [[START]], i32* [[START_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[END]], i32* [[END_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[STEP]], i32* [[STEP_ADDR]], align 4
// CHECK1-NEXT:    store i32 7, i32* [[I]], align 4
// CHECK1-NEXT:    store i32 7, i32* [[J]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
// CHECK1:       for.cond:
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], 4
// CHECK1-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END30:%.*]]
// CHECK1:       for.body:
// CHECK1-NEXT:    store i32 0, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND1:%.*]]
// CHECK1:       for.cond1:
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP1]], 4
// CHECK1-NEXT:    br i1 [[CMP2]], label [[FOR_BODY3:%.*]], label [[FOR_END27:%.*]]
// CHECK1:       for.body3:
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND4:%.*]]
// CHECK1:       for.cond4:
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP4]], 5
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp slt i32 4, [[ADD]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP5]], 5
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[ADD6]], [[COND_FALSE]] ]
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp slt i32 [[TMP3]], [[COND]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[FOR_BODY8:%.*]], label [[FOR_END24:%.*]]
// CHECK1:       for.body8:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 3
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 7, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND10:%.*]]
// CHECK1:       for.cond10:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP9]], 5
// CHECK1-NEXT:    [[CMP12:%.*]] = icmp slt i32 4, [[ADD11]]
// CHECK1-NEXT:    br i1 [[CMP12]], label [[COND_TRUE13:%.*]], label [[COND_FALSE14:%.*]]
// CHECK1:       cond.true13:
// CHECK1-NEXT:    br label [[COND_END16:%.*]]
// CHECK1:       cond.false14:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK1-NEXT:    br label [[COND_END16]]
// CHECK1:       cond.end16:
// CHECK1-NEXT:    [[COND17:%.*]] = phi i32 [ 4, [[COND_TRUE13]] ], [ [[ADD15]], [[COND_FALSE14]] ]
// CHECK1-NEXT:    [[CMP18:%.*]] = icmp slt i32 [[TMP8]], [[COND17]]
// CHECK1-NEXT:    br i1 [[CMP18]], label [[FOR_BODY19:%.*]], label [[FOR_END:%.*]]
// CHECK1:       for.body19:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[MUL20:%.*]] = mul nsw i32 [[TMP11]], 3
// CHECK1-NEXT:    [[ADD21:%.*]] = add nsw i32 7, [[MUL20]]
// CHECK1-NEXT:    store i32 [[ADD21]], i32* [[J]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[J]], align 4
// CHECK1-NEXT:    call void (...) @body(i32 [[TMP12]], i32 [[TMP13]])
// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
// CHECK1:       for.inc:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND10]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK1:       for.end:
// CHECK1-NEXT:    br label [[FOR_INC22:%.*]]
// CHECK1:       for.inc22:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[INC23:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK1-NEXT:    store i32 [[INC23]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND4]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK1:       for.end24:
// CHECK1-NEXT:    br label [[FOR_INC25:%.*]]
// CHECK1:       for.inc25:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP16]], 5
// CHECK1-NEXT:    store i32 [[ADD26]], i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND1]], !llvm.loop [[LOOP9:![0-9]+]]
// CHECK1:       for.end27:
// CHECK1-NEXT:    br label [[FOR_INC28:%.*]]
// CHECK1:       for.inc28:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP17]], 5
// CHECK1-NEXT:    store i32 [[ADD29]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK1:       for.end30:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@foo3
// CHECK1-SAME: () #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_1_IV_J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_1_IV_J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
// CHECK1-NEXT:    store i32 7, i32* [[I]], align 4
// CHECK1-NEXT:    store i32 7, i32* [[J]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 0, 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* @[[GLOB1:[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:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 0
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 0, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 5
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
// CHECK1:       for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP7]], 4
// CHECK1-NEXT:    br i1 [[CMP2]], label [[FOR_BODY:%.*]], label [[FOR_END32:%.*]]
// CHECK1:       for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK1:       for.cond3:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp slt i32 4, [[ADD4]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK1:       cond.true6:
// CHECK1-NEXT:    br label [[COND_END9:%.*]]
// CHECK1:       cond.false7:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP11]], 5
// CHECK1-NEXT:    br label [[COND_END9]]
// CHECK1:       cond.end9:
// CHECK1-NEXT:    [[COND10:%.*]] = phi i32 [ 4, [[COND_TRUE6]] ], [ [[ADD8]], [[COND_FALSE7]] ]
// CHECK1-NEXT:    [[CMP11:%.*]] = icmp slt i32 [[TMP9]], [[COND10]]
// CHECK1-NEXT:    br i1 [[CMP11]], label [[FOR_BODY12:%.*]], label [[FOR_END29:%.*]]
// CHECK1:       for.body12:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP12]], 3
// CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 7, [[MUL13]]
// CHECK1-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND15:%.*]]
// CHECK1:       for.cond15:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP15]], 5
// CHECK1-NEXT:    [[CMP17:%.*]] = icmp slt i32 4, [[ADD16]]
// CHECK1-NEXT:    br i1 [[CMP17]], label [[COND_TRUE18:%.*]], label [[COND_FALSE19:%.*]]
// CHECK1:       cond.true18:
// CHECK1-NEXT:    br label [[COND_END21:%.*]]
// CHECK1:       cond.false19:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP16]], 5
// CHECK1-NEXT:    br label [[COND_END21]]
// CHECK1:       cond.end21:
// CHECK1-NEXT:    [[COND22:%.*]] = phi i32 [ 4, [[COND_TRUE18]] ], [ [[ADD20]], [[COND_FALSE19]] ]
// CHECK1-NEXT:    [[CMP23:%.*]] = icmp slt i32 [[TMP14]], [[COND22]]
// CHECK1-NEXT:    br i1 [[CMP23]], label [[FOR_BODY24:%.*]], label [[FOR_END:%.*]]
// CHECK1:       for.body24:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP17]], 3
// CHECK1-NEXT:    [[ADD26:%.*]] = add nsw i32 7, [[MUL25]]
// CHECK1-NEXT:    store i32 [[ADD26]], i32* [[J]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[J]], align 4
// CHECK1-NEXT:    call void (...) @body(i32 [[TMP18]], i32 [[TMP19]])
// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
// CHECK1:       for.inc:
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND15]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK1:       for.end:
// CHECK1-NEXT:    br label [[FOR_INC27:%.*]]
// CHECK1:       for.inc27:
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[INC28:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK1-NEXT:    store i32 [[INC28]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP12:![0-9]+]]
// CHECK1:       for.end29:
// CHECK1-NEXT:    br label [[FOR_INC30:%.*]]
// CHECK1:       for.inc30:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP22]], 5
// CHECK1-NEXT:    store i32 [[ADD31]], i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK1:       for.end32:
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD33:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK1-NEXT:    store i32 [[ADD33]], 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:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@foo4
// CHECK1-SAME: () #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[K:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_1_IV_J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_1_IV_J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
// CHECK1-NEXT:    store i32 7, i32* [[I]], align 4
// CHECK1-NEXT:    store i32 7, i32* [[J]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 3, 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* @[[GLOB1]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 3
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP6]], 1
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 3
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 7, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[K]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP8]], 1
// CHECK1-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 1
// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], [[MUL4]]
// CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 5
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
// CHECK1:       for.cond:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp slt i32 [[TMP9]], 4
// CHECK1-NEXT:    br i1 [[CMP7]], label [[FOR_BODY:%.*]], label [[FOR_END37:%.*]]
// CHECK1:       for.body:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 [[TMP10]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND8:%.*]]
// CHECK1:       for.cond8:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP12]], 5
// CHECK1-NEXT:    [[CMP10:%.*]] = icmp slt i32 4, [[ADD9]]
// CHECK1-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK1:       cond.true11:
// CHECK1-NEXT:    br label [[COND_END14:%.*]]
// CHECK1:       cond.false12:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP13]], 5
// CHECK1-NEXT:    br label [[COND_END14]]
// CHECK1:       cond.end14:
// CHECK1-NEXT:    [[COND15:%.*]] = phi i32 [ 4, [[COND_TRUE11]] ], [ [[ADD13]], [[COND_FALSE12]] ]
// CHECK1-NEXT:    [[CMP16:%.*]] = icmp slt i32 [[TMP11]], [[COND15]]
// CHECK1-NEXT:    br i1 [[CMP16]], label [[FOR_BODY17:%.*]], label [[FOR_END34:%.*]]
// CHECK1:       for.body17:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[MUL18:%.*]] = mul nsw i32 [[TMP14]], 3
// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 7, [[MUL18]]
// CHECK1-NEXT:    store i32 [[ADD19]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    store i32 [[TMP15]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND20:%.*]]
// CHECK1:       for.cond20:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP17]], 5
// CHECK1-NEXT:    [[CMP22:%.*]] = icmp slt i32 4, [[ADD21]]
// CHECK1-NEXT:    br i1 [[CMP22]], label [[COND_TRUE23:%.*]], label [[COND_FALSE24:%.*]]
// CHECK1:       cond.true23:
// CHECK1-NEXT:    br label [[COND_END26:%.*]]
// CHECK1:       cond.false24:
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP18]], 5
// CHECK1-NEXT:    br label [[COND_END26]]
// CHECK1:       cond.end26:
// CHECK1-NEXT:    [[COND27:%.*]] = phi i32 [ 4, [[COND_TRUE23]] ], [ [[ADD25]], [[COND_FALSE24]] ]
// CHECK1-NEXT:    [[CMP28:%.*]] = icmp slt i32 [[TMP16]], [[COND27]]
// CHECK1-NEXT:    br i1 [[CMP28]], label [[FOR_BODY29:%.*]], label [[FOR_END:%.*]]
// CHECK1:       for.body29:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[MUL30:%.*]] = mul nsw i32 [[TMP19]], 3
// CHECK1-NEXT:    [[ADD31:%.*]] = add nsw i32 7, [[MUL30]]
// CHECK1-NEXT:    store i32 [[ADD31]], i32* [[J]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[J]], align 4
// CHECK1-NEXT:    call void (...) @body(i32 [[TMP20]], i32 [[TMP21]])
// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
// CHECK1:       for.inc:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND20]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK1:       for.end:
// CHECK1-NEXT:    br label [[FOR_INC32:%.*]]
// CHECK1:       for.inc32:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[INC33:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK1-NEXT:    store i32 [[INC33]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND8]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK1:       for.end34:
// CHECK1-NEXT:    br label [[FOR_INC35:%.*]]
// CHECK1:       for.inc35:
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    [[ADD36:%.*]] = add nsw i32 [[TMP24]], 5
// CHECK1-NEXT:    store i32 [[ADD36]], i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK1:       for.end37:
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP25]], 1
// CHECK1-NEXT:    store i32 [[ADD38]], 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:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@foo5
// CHECK1-SAME: () #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I11:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I12:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[J13:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
// CHECK1-NEXT:    store i32 7, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP2]], 5
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 4, [[ADD]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP]], align 4
// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 5
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[ADD4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], [[TMP5]]
// CHECK1-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
// CHECK1-NEXT:    [[ADD7:%.*]] = add i32 [[SUB6]], 1
// CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD7]], 1
// CHECK1-NEXT:    [[CONV:%.*]] = zext i32 [[DIV]] to i64
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i64 1, [[CONV]]
// CHECK1-NEXT:    [[MUL8:%.*]] = mul nsw i64 [[MUL]], 4
// CHECK1-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL8]], 1
// CHECK1-NEXT:    store i64 [[SUB9]], i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 7, i32* [[J]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[CMP10:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK1:       omp.precond.then:
// CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
// CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK1-NEXT:    [[CMP14:%.*]] = icmp sgt i64 [[TMP10]], [[TMP11]]
// CHECK1-NEXT:    br i1 [[CMP14]], label [[COND_TRUE15:%.*]], label [[COND_FALSE16:%.*]]
// CHECK1:       cond.true15:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK1-NEXT:    br label [[COND_END17:%.*]]
// CHECK1:       cond.false16:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    br label [[COND_END17]]
// CHECK1:       cond.end17:
// CHECK1-NEXT:    [[COND18:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE15]] ], [ [[TMP13]], [[COND_FALSE16]] ]
// CHECK1-NEXT:    store i64 [[COND18]], i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
// CHECK1-NEXT:    store i64 [[TMP14]], i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK1-NEXT:    [[CMP19:%.*]] = icmp sle i64 [[TMP15]], [[TMP16]]
// CHECK1-NEXT:    br i1 [[CMP19]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB20:%.*]] = sub i32 [[TMP18]], [[TMP19]]
// CHECK1-NEXT:    [[SUB21:%.*]] = sub i32 [[SUB20]], 1
// CHECK1-NEXT:    [[ADD22:%.*]] = add i32 [[SUB21]], 1
// CHECK1-NEXT:    [[DIV23:%.*]] = udiv i32 [[ADD22]], 1
// CHECK1-NEXT:    [[MUL24:%.*]] = mul i32 1, [[DIV23]]
// CHECK1-NEXT:    [[MUL25:%.*]] = mul i32 [[MUL24]], 4
// CHECK1-NEXT:    [[CONV26:%.*]] = zext i32 [[MUL25]] to i64
// CHECK1-NEXT:    [[DIV27:%.*]] = sdiv i64 [[TMP17]], [[CONV26]]
// CHECK1-NEXT:    [[MUL28:%.*]] = mul nsw i64 [[DIV27]], 5
// CHECK1-NEXT:    [[ADD29:%.*]] = add nsw i64 0, [[MUL28]]
// CHECK1-NEXT:    [[CONV30:%.*]] = trunc i64 [[ADD29]] to i32
// CHECK1-NEXT:    store i32 [[CONV30]], i32* [[DOTFLOOR_0_IV_I11]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CONV31:%.*]] = sext i32 [[TMP20]] to i64
// CHECK1-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB32:%.*]] = sub i32 [[TMP23]], [[TMP24]]
// CHECK1-NEXT:    [[SUB33:%.*]] = sub i32 [[SUB32]], 1
// CHECK1-NEXT:    [[ADD34:%.*]] = add i32 [[SUB33]], 1
// CHECK1-NEXT:    [[DIV35:%.*]] = udiv i32 [[ADD34]], 1
// CHECK1-NEXT:    [[MUL36:%.*]] = mul i32 1, [[DIV35]]
// CHECK1-NEXT:    [[MUL37:%.*]] = mul i32 [[MUL36]], 4
// CHECK1-NEXT:    [[CONV38:%.*]] = zext i32 [[MUL37]] to i64
// CHECK1-NEXT:    [[DIV39:%.*]] = sdiv i64 [[TMP22]], [[CONV38]]
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB40:%.*]] = sub i32 [[TMP25]], [[TMP26]]
// CHECK1-NEXT:    [[SUB41:%.*]] = sub i32 [[SUB40]], 1
// CHECK1-NEXT:    [[ADD42:%.*]] = add i32 [[SUB41]], 1
// CHECK1-NEXT:    [[DIV43:%.*]] = udiv i32 [[ADD42]], 1
// CHECK1-NEXT:    [[MUL44:%.*]] = mul i32 1, [[DIV43]]
// CHECK1-NEXT:    [[MUL45:%.*]] = mul i32 [[MUL44]], 4
// CHECK1-NEXT:    [[CONV46:%.*]] = zext i32 [[MUL45]] to i64
// CHECK1-NEXT:    [[MUL47:%.*]] = mul nsw i64 [[DIV39]], [[CONV46]]
// CHECK1-NEXT:    [[SUB48:%.*]] = sub nsw i64 [[TMP21]], [[MUL47]]
// CHECK1-NEXT:    [[DIV49:%.*]] = sdiv i64 [[SUB48]], 4
// CHECK1-NEXT:    [[MUL50:%.*]] = mul nsw i64 [[DIV49]], 1
// CHECK1-NEXT:    [[ADD51:%.*]] = add nsw i64 [[CONV31]], [[MUL50]]
// CHECK1-NEXT:    [[CONV52:%.*]] = trunc i64 [[ADD51]] to i32
// CHECK1-NEXT:    store i32 [[CONV52]], i32* [[DOTTILE_0_IV_I12]], align 4
// CHECK1-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB53:%.*]] = sub i32 [[TMP29]], [[TMP30]]
// CHECK1-NEXT:    [[SUB54:%.*]] = sub i32 [[SUB53]], 1
// CHECK1-NEXT:    [[ADD55:%.*]] = add i32 [[SUB54]], 1
// CHECK1-NEXT:    [[DIV56:%.*]] = udiv i32 [[ADD55]], 1
// CHECK1-NEXT:    [[MUL57:%.*]] = mul i32 1, [[DIV56]]
// CHECK1-NEXT:    [[MUL58:%.*]] = mul i32 [[MUL57]], 4
// CHECK1-NEXT:    [[CONV59:%.*]] = zext i32 [[MUL58]] to i64
// CHECK1-NEXT:    [[DIV60:%.*]] = sdiv i64 [[TMP28]], [[CONV59]]
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB61:%.*]] = sub i32 [[TMP31]], [[TMP32]]
// CHECK1-NEXT:    [[SUB62:%.*]] = sub i32 [[SUB61]], 1
// CHECK1-NEXT:    [[ADD63:%.*]] = add i32 [[SUB62]], 1
// CHECK1-NEXT:    [[DIV64:%.*]] = udiv i32 [[ADD63]], 1
// CHECK1-NEXT:    [[MUL65:%.*]] = mul i32 1, [[DIV64]]
// CHECK1-NEXT:    [[MUL66:%.*]] = mul i32 [[MUL65]], 4
// CHECK1-NEXT:    [[CONV67:%.*]] = zext i32 [[MUL66]] to i64
// CHECK1-NEXT:    [[MUL68:%.*]] = mul nsw i64 [[DIV60]], [[CONV67]]
// CHECK1-NEXT:    [[SUB69:%.*]] = sub nsw i64 [[TMP27]], [[MUL68]]
// CHECK1-NEXT:    [[TMP33:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB70:%.*]] = sub i32 [[TMP35]], [[TMP36]]
// CHECK1-NEXT:    [[SUB71:%.*]] = sub i32 [[SUB70]], 1
// CHECK1-NEXT:    [[ADD72:%.*]] = add i32 [[SUB71]], 1
// CHECK1-NEXT:    [[DIV73:%.*]] = udiv i32 [[ADD72]], 1
// CHECK1-NEXT:    [[MUL74:%.*]] = mul i32 1, [[DIV73]]
// CHECK1-NEXT:    [[MUL75:%.*]] = mul i32 [[MUL74]], 4
// CHECK1-NEXT:    [[CONV76:%.*]] = zext i32 [[MUL75]] to i64
// CHECK1-NEXT:    [[DIV77:%.*]] = sdiv i64 [[TMP34]], [[CONV76]]
// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB78:%.*]] = sub i32 [[TMP37]], [[TMP38]]
// CHECK1-NEXT:    [[SUB79:%.*]] = sub i32 [[SUB78]], 1
// CHECK1-NEXT:    [[ADD80:%.*]] = add i32 [[SUB79]], 1
// CHECK1-NEXT:    [[DIV81:%.*]] = udiv i32 [[ADD80]], 1
// CHECK1-NEXT:    [[MUL82:%.*]] = mul i32 1, [[DIV81]]
// CHECK1-NEXT:    [[MUL83:%.*]] = mul i32 [[MUL82]], 4
// CHECK1-NEXT:    [[CONV84:%.*]] = zext i32 [[MUL83]] to i64
// CHECK1-NEXT:    [[MUL85:%.*]] = mul nsw i64 [[DIV77]], [[CONV84]]
// CHECK1-NEXT:    [[SUB86:%.*]] = sub nsw i64 [[TMP33]], [[MUL85]]
// CHECK1-NEXT:    [[DIV87:%.*]] = sdiv i64 [[SUB86]], 4
// CHECK1-NEXT:    [[MUL88:%.*]] = mul nsw i64 [[DIV87]], 4
// CHECK1-NEXT:    [[SUB89:%.*]] = sub nsw i64 [[SUB69]], [[MUL88]]
// CHECK1-NEXT:    [[MUL90:%.*]] = mul nsw i64 [[SUB89]], 3
// CHECK1-NEXT:    [[ADD91:%.*]] = add nsw i64 7, [[MUL90]]
// CHECK1-NEXT:    [[CONV92:%.*]] = trunc i64 [[ADD91]] to i32
// CHECK1-NEXT:    store i32 [[CONV92]], i32* [[J13]], align 4
// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTTILE_0_IV_I12]], align 4
// CHECK1-NEXT:    [[MUL93:%.*]] = mul nsw i32 [[TMP39]], 3
// CHECK1-NEXT:    [[ADD94:%.*]] = add nsw i32 7, [[MUL93]]
// CHECK1-NEXT:    store i32 [[ADD94]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[J13]], align 4
// CHECK1-NEXT:    call void (...) @body(i32 [[TMP40]], i32 [[TMP41]])
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP42:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK1-NEXT:    [[ADD95:%.*]] = add nsw i64 [[TMP42]], 1
// CHECK1-NEXT:    store i64 [[ADD95]], i64* [[DOTOMP_IV]], align 8
// 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:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       omp.precond.end:
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@foo6
// CHECK1-SAME: () #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I:%.*]] = 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 7, i32* [[I]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 0, 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:    [[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* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 0
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 0, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
// CHECK1:       for.cond:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK1-NEXT:    [[CMP3:%.*]] = icmp slt i32 4, [[ADD2]]
// CHECK1-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
// CHECK1:       cond.true4:
// CHECK1-NEXT:    br label [[COND_END7:%.*]]
// CHECK1:       cond.false5:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 5
// CHECK1-NEXT:    br label [[COND_END7]]
// CHECK1:       cond.end7:
// CHECK1-NEXT:    [[COND8:%.*]] = phi i32 [ 4, [[COND_TRUE4]] ], [ [[ADD6]], [[COND_FALSE5]] ]
// CHECK1-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP9]], [[COND8]]
// CHECK1-NEXT:    br i1 [[CMP9]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK1:       for.body:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP12]], 3
// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 7, [[MUL10]]
// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    call void (...) @body(i32 [[TMP13]])
// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
// CHECK1:       for.inc:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK1:       for.end:
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK1-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@tfoo7
// CHECK1-SAME: () #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @_Z4foo7IiLi3ELi5EEvT_S0_(i32 0, i32 42)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z4foo7IiLi3ELi5EEvT_S0_
// CHECK1-SAME: (i32 [[START:%.*]], i32 [[END:%.*]]) #[[ATTR0]] comdat {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[START_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[END_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32 [[START]], i32* [[START_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[END]], i32* [[END_ADDR]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[START_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[START_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[END_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
// CHECK1-NEXT:    [[SUB3:%.*]] = sub i32 [[SUB]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add i32 [[SUB3]], 3
// CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 3
// CHECK1-NEXT:    [[SUB4:%.*]] = sub i32 [[DIV]], 1
// CHECK1-NEXT:    store i32 [[SUB4]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
// CHECK1:       for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[ADD5:%.*]] = add i32 [[TMP6]], 1
// CHECK1-NEXT:    [[CMP:%.*]] = icmp ult i32 [[TMP5]], [[ADD5]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END17:%.*]]
// CHECK1:       for.body:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND6:%.*]]
// CHECK1:       for.cond6:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add i32 [[TMP9]], 1
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK1-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[ADD7]], [[ADD8]]
// CHECK1-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[ADD10:%.*]] = add i32 [[TMP11]], 1
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP12]], 5
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[ADD10]], [[COND_TRUE]] ], [ [[ADD11]], [[COND_FALSE]] ]
// CHECK1-NEXT:    [[CMP12:%.*]] = icmp ult i32 [[TMP8]], [[COND]]
// CHECK1-NEXT:    br i1 [[CMP12]], label [[FOR_BODY13:%.*]], label [[FOR_END:%.*]]
// CHECK1:       for.body13:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul i32 [[TMP14]], 3
// CHECK1-NEXT:    [[ADD14:%.*]] = add i32 [[TMP13]], [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    call void (...) @body(i32 [[TMP15]])
// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
// CHECK1:       for.inc:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND6]], !llvm.loop [[LOOP20:![0-9]+]]
// CHECK1:       for.end:
// CHECK1-NEXT:    br label [[FOR_INC15:%.*]]
// CHECK1:       for.inc15:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP17]], 5
// CHECK1-NEXT:    store i32 [[ADD16]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
// CHECK1:       for.end17:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_tile_codegen.cpp
// CHECK1-SAME: () #[[ATTR1]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__cxx_global_var_init()
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@__cxx_global_var_init
// CHECK2-SAME: () #[[ATTR0:[0-9]+]] section ".text.startup" {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @_ZN1SC1Ev(%struct.S* nonnull align 4 dereferenceable(4) @s)
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SC1Ev
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void @_ZN1SC2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SC2Ev
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[I2:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    store i32* [[I2]], i32** [[I]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], 4
// CHECK2-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END11:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK2:       for.cond3:
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 5
// CHECK2-NEXT:    [[CMP4:%.*]] = icmp slt i32 4, [[ADD]]
// CHECK2-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP4]], 5
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[ADD5]], [[COND_FALSE]] ]
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp slt i32 [[TMP2]], [[COND]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[FOR_BODY7:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body7:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP5]], 3
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 7, [[MUL]]
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[I]], align 8
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[TMP6]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[I]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK2-NEXT:    call void (...) @body(i32 [[TMP8]])
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    br label [[FOR_INC9:%.*]]
// CHECK2:       for.inc9:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK2-NEXT:    store i32 [[ADD10]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK2:       for.end11:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@body
// CHECK2-SAME: (...) #[[ATTR2:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@foo1
// CHECK2-SAME: (i32 [[START:%.*]], i32 [[END:%.*]], i32 [[STEP:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[START_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[END_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[STEP_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32 [[START]], i32* [[START_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[END]], i32* [[END_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[STEP]], i32* [[STEP_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[START_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[END_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[STEP_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
// CHECK2-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], [[TMP5]]
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], [[TMP6]]
// CHECK2-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[ADD6:%.*]] = add i32 [[TMP8]], 1
// CHECK2-NEXT:    [[CMP:%.*]] = icmp ult i32 [[TMP7]], [[ADD6]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END18:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND7:%.*]]
// CHECK2:       for.cond7:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add i32 [[TMP11]], 1
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP12]], 5
// CHECK2-NEXT:    [[CMP10:%.*]] = icmp ult i32 [[ADD8]], [[ADD9]]
// CHECK2-NEXT:    br i1 [[CMP10]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[ADD11:%.*]] = add i32 [[TMP13]], 1
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP14]], 5
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[ADD11]], [[COND_TRUE]] ], [ [[ADD12]], [[COND_FALSE]] ]
// CHECK2-NEXT:    [[CMP13:%.*]] = icmp ult i32 [[TMP10]], [[COND]]
// CHECK2-NEXT:    br i1 [[CMP13]], label [[FOR_BODY14:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body14:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], [[TMP17]]
// CHECK2-NEXT:    [[ADD15:%.*]] = add i32 [[TMP15]], [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD15]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    call void (...) @body(i32 [[TMP18]])
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP19]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND7]], !llvm.loop [[LOOP5:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    br label [[FOR_INC16:%.*]]
// CHECK2:       for.inc16:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 5
// CHECK2-NEXT:    store i32 [[ADD17]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
// CHECK2:       for.end18:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@foo2
// CHECK2-SAME: (i32 [[START:%.*]], i32 [[END:%.*]], i32 [[STEP:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[START_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[END_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[STEP_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_1_IV_J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_1_IV_J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32 [[START]], i32* [[START_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[END]], i32* [[END_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[STEP]], i32* [[STEP_ADDR]], align 4
// CHECK2-NEXT:    store i32 7, i32* [[I]], align 4
// CHECK2-NEXT:    store i32 7, i32* [[J]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], 4
// CHECK2-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END30:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    store i32 0, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND1:%.*]]
// CHECK2:       for.cond1:
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP1]], 4
// CHECK2-NEXT:    br i1 [[CMP2]], label [[FOR_BODY3:%.*]], label [[FOR_END27:%.*]]
// CHECK2:       for.body3:
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND4:%.*]]
// CHECK2:       for.cond4:
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP4]], 5
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp slt i32 4, [[ADD]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP5]], 5
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[ADD6]], [[COND_FALSE]] ]
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp slt i32 [[TMP3]], [[COND]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[FOR_BODY8:%.*]], label [[FOR_END24:%.*]]
// CHECK2:       for.body8:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 3
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 7, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND10:%.*]]
// CHECK2:       for.cond10:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP9]], 5
// CHECK2-NEXT:    [[CMP12:%.*]] = icmp slt i32 4, [[ADD11]]
// CHECK2-NEXT:    br i1 [[CMP12]], label [[COND_TRUE13:%.*]], label [[COND_FALSE14:%.*]]
// CHECK2:       cond.true13:
// CHECK2-NEXT:    br label [[COND_END16:%.*]]
// CHECK2:       cond.false14:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK2-NEXT:    br label [[COND_END16]]
// CHECK2:       cond.end16:
// CHECK2-NEXT:    [[COND17:%.*]] = phi i32 [ 4, [[COND_TRUE13]] ], [ [[ADD15]], [[COND_FALSE14]] ]
// CHECK2-NEXT:    [[CMP18:%.*]] = icmp slt i32 [[TMP8]], [[COND17]]
// CHECK2-NEXT:    br i1 [[CMP18]], label [[FOR_BODY19:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body19:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[MUL20:%.*]] = mul nsw i32 [[TMP11]], 3
// CHECK2-NEXT:    [[ADD21:%.*]] = add nsw i32 7, [[MUL20]]
// CHECK2-NEXT:    store i32 [[ADD21]], i32* [[J]], align 4
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[J]], align 4
// CHECK2-NEXT:    call void (...) @body(i32 [[TMP12]], i32 [[TMP13]])
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND10]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    br label [[FOR_INC22:%.*]]
// CHECK2:       for.inc22:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[INC23:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK2-NEXT:    store i32 [[INC23]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND4]], !llvm.loop [[LOOP8:![0-9]+]]
// CHECK2:       for.end24:
// CHECK2-NEXT:    br label [[FOR_INC25:%.*]]
// CHECK2:       for.inc25:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP16]], 5
// CHECK2-NEXT:    store i32 [[ADD26]], i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND1]], !llvm.loop [[LOOP9:![0-9]+]]
// CHECK2:       for.end27:
// CHECK2-NEXT:    br label [[FOR_INC28:%.*]]
// CHECK2:       for.inc28:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP17]], 5
// CHECK2-NEXT:    store i32 [[ADD29]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK2:       for.end30:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@foo3
// CHECK2-SAME: () #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_1_IV_J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_1_IV_J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
// CHECK2-NEXT:    store i32 7, i32* [[I]], align 4
// CHECK2-NEXT:    store i32 7, i32* [[J]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 0, 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* @[[GLOB1:[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:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 0
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 0, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 5
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[TMP7]], 4
// CHECK2-NEXT:    br i1 [[CMP2]], label [[FOR_BODY:%.*]], label [[FOR_END32:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK2:       for.cond3:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp slt i32 4, [[ADD4]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
// CHECK2:       cond.true6:
// CHECK2-NEXT:    br label [[COND_END9:%.*]]
// CHECK2:       cond.false7:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP11]], 5
// CHECK2-NEXT:    br label [[COND_END9]]
// CHECK2:       cond.end9:
// CHECK2-NEXT:    [[COND10:%.*]] = phi i32 [ 4, [[COND_TRUE6]] ], [ [[ADD8]], [[COND_FALSE7]] ]
// CHECK2-NEXT:    [[CMP11:%.*]] = icmp slt i32 [[TMP9]], [[COND10]]
// CHECK2-NEXT:    br i1 [[CMP11]], label [[FOR_BODY12:%.*]], label [[FOR_END29:%.*]]
// CHECK2:       for.body12:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP12]], 3
// CHECK2-NEXT:    [[ADD14:%.*]] = add nsw i32 7, [[MUL13]]
// CHECK2-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND15:%.*]]
// CHECK2:       for.cond15:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP15]], 5
// CHECK2-NEXT:    [[CMP17:%.*]] = icmp slt i32 4, [[ADD16]]
// CHECK2-NEXT:    br i1 [[CMP17]], label [[COND_TRUE18:%.*]], label [[COND_FALSE19:%.*]]
// CHECK2:       cond.true18:
// CHECK2-NEXT:    br label [[COND_END21:%.*]]
// CHECK2:       cond.false19:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP16]], 5
// CHECK2-NEXT:    br label [[COND_END21]]
// CHECK2:       cond.end21:
// CHECK2-NEXT:    [[COND22:%.*]] = phi i32 [ 4, [[COND_TRUE18]] ], [ [[ADD20]], [[COND_FALSE19]] ]
// CHECK2-NEXT:    [[CMP23:%.*]] = icmp slt i32 [[TMP14]], [[COND22]]
// CHECK2-NEXT:    br i1 [[CMP23]], label [[FOR_BODY24:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body24:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP17]], 3
// CHECK2-NEXT:    [[ADD26:%.*]] = add nsw i32 7, [[MUL25]]
// CHECK2-NEXT:    store i32 [[ADD26]], i32* [[J]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[J]], align 4
// CHECK2-NEXT:    call void (...) @body(i32 [[TMP18]], i32 [[TMP19]])
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND15]], !llvm.loop [[LOOP11:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    br label [[FOR_INC27:%.*]]
// CHECK2:       for.inc27:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[INC28:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK2-NEXT:    store i32 [[INC28]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP12:![0-9]+]]
// CHECK2:       for.end29:
// CHECK2-NEXT:    br label [[FOR_INC30:%.*]]
// CHECK2:       for.inc30:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD31:%.*]] = add nsw i32 [[TMP22]], 5
// CHECK2-NEXT:    store i32 [[ADD31]], i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK2:       for.end32:
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD33:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK2-NEXT:    store i32 [[ADD33]], 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:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@foo4
// CHECK2-SAME: () #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[K:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_1_IV_J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_1_IV_J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
// CHECK2-NEXT:    store i32 7, i32* [[I]], align 4
// CHECK2-NEXT:    store i32 7, i32* [[J]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 3, 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* @[[GLOB1]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 3
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
// CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP6]], 1
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 3
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 7, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[K]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP8]], 1
// CHECK2-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 1
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], [[MUL4]]
// CHECK2-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 5
// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp slt i32 [[TMP9]], 4
// CHECK2-NEXT:    br i1 [[CMP7]], label [[FOR_BODY:%.*]], label [[FOR_END37:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 [[TMP10]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND8:%.*]]
// CHECK2:       for.cond8:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP12]], 5
// CHECK2-NEXT:    [[CMP10:%.*]] = icmp slt i32 4, [[ADD9]]
// CHECK2-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
// CHECK2:       cond.true11:
// CHECK2-NEXT:    br label [[COND_END14:%.*]]
// CHECK2:       cond.false12:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP13]], 5
// CHECK2-NEXT:    br label [[COND_END14]]
// CHECK2:       cond.end14:
// CHECK2-NEXT:    [[COND15:%.*]] = phi i32 [ 4, [[COND_TRUE11]] ], [ [[ADD13]], [[COND_FALSE12]] ]
// CHECK2-NEXT:    [[CMP16:%.*]] = icmp slt i32 [[TMP11]], [[COND15]]
// CHECK2-NEXT:    br i1 [[CMP16]], label [[FOR_BODY17:%.*]], label [[FOR_END34:%.*]]
// CHECK2:       for.body17:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[MUL18:%.*]] = mul nsw i32 [[TMP14]], 3
// CHECK2-NEXT:    [[ADD19:%.*]] = add nsw i32 7, [[MUL18]]
// CHECK2-NEXT:    store i32 [[ADD19]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    store i32 [[TMP15]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND20:%.*]]
// CHECK2:       for.cond20:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP17]], 5
// CHECK2-NEXT:    [[CMP22:%.*]] = icmp slt i32 4, [[ADD21]]
// CHECK2-NEXT:    br i1 [[CMP22]], label [[COND_TRUE23:%.*]], label [[COND_FALSE24:%.*]]
// CHECK2:       cond.true23:
// CHECK2-NEXT:    br label [[COND_END26:%.*]]
// CHECK2:       cond.false24:
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP18]], 5
// CHECK2-NEXT:    br label [[COND_END26]]
// CHECK2:       cond.end26:
// CHECK2-NEXT:    [[COND27:%.*]] = phi i32 [ 4, [[COND_TRUE23]] ], [ [[ADD25]], [[COND_FALSE24]] ]
// CHECK2-NEXT:    [[CMP28:%.*]] = icmp slt i32 [[TMP16]], [[COND27]]
// CHECK2-NEXT:    br i1 [[CMP28]], label [[FOR_BODY29:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body29:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[MUL30:%.*]] = mul nsw i32 [[TMP19]], 3
// CHECK2-NEXT:    [[ADD31:%.*]] = add nsw i32 7, [[MUL30]]
// CHECK2-NEXT:    store i32 [[ADD31]], i32* [[J]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[J]], align 4
// CHECK2-NEXT:    call void (...) @body(i32 [[TMP20]], i32 [[TMP21]])
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTTILE_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND20]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    br label [[FOR_INC32:%.*]]
// CHECK2:       for.inc32:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[INC33:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK2-NEXT:    store i32 [[INC33]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND8]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK2:       for.end34:
// CHECK2-NEXT:    br label [[FOR_INC35:%.*]]
// CHECK2:       for.inc35:
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    [[ADD36:%.*]] = add nsw i32 [[TMP24]], 5
// CHECK2-NEXT:    store i32 [[ADD36]], i32* [[DOTFLOOR_1_IV_J]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK2:       for.end37:
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP25]], 1
// CHECK2-NEXT:    store i32 [[ADD38]], 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:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@foo5
// CHECK2-SAME: () #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I11:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I12:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[J13:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
// CHECK2-NEXT:    store i32 7, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP2]], 5
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 4, [[ADD]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP]], align 4
// CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 5
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[ADD4]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], [[TMP5]]
// CHECK2-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
// CHECK2-NEXT:    [[ADD7:%.*]] = add i32 [[SUB6]], 1
// CHECK2-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD7]], 1
// CHECK2-NEXT:    [[CONV:%.*]] = zext i32 [[DIV]] to i64
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i64 1, [[CONV]]
// CHECK2-NEXT:    [[MUL8:%.*]] = mul nsw i64 [[MUL]], 4
// CHECK2-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL8]], 1
// CHECK2-NEXT:    store i64 [[SUB9]], i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 7, i32* [[J]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[CMP10:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK2-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
// CHECK2-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_UB]], align 8
// CHECK2-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
// CHECK2-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK2-NEXT:    [[CMP14:%.*]] = icmp sgt i64 [[TMP10]], [[TMP11]]
// CHECK2-NEXT:    br i1 [[CMP14]], label [[COND_TRUE15:%.*]], label [[COND_FALSE16:%.*]]
// CHECK2:       cond.true15:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
// CHECK2-NEXT:    br label [[COND_END17:%.*]]
// CHECK2:       cond.false16:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK2-NEXT:    br label [[COND_END17]]
// CHECK2:       cond.end17:
// CHECK2-NEXT:    [[COND18:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE15]] ], [ [[TMP13]], [[COND_FALSE16]] ]
// CHECK2-NEXT:    store i64 [[COND18]], i64* [[DOTOMP_UB]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
// CHECK2-NEXT:    store i64 [[TMP14]], i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
// CHECK2-NEXT:    [[CMP19:%.*]] = icmp sle i64 [[TMP15]], [[TMP16]]
// CHECK2-NEXT:    br i1 [[CMP19]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB20:%.*]] = sub i32 [[TMP18]], [[TMP19]]
// CHECK2-NEXT:    [[SUB21:%.*]] = sub i32 [[SUB20]], 1
// CHECK2-NEXT:    [[ADD22:%.*]] = add i32 [[SUB21]], 1
// CHECK2-NEXT:    [[DIV23:%.*]] = udiv i32 [[ADD22]], 1
// CHECK2-NEXT:    [[MUL24:%.*]] = mul i32 1, [[DIV23]]
// CHECK2-NEXT:    [[MUL25:%.*]] = mul i32 [[MUL24]], 4
// CHECK2-NEXT:    [[CONV26:%.*]] = zext i32 [[MUL25]] to i64
// CHECK2-NEXT:    [[DIV27:%.*]] = sdiv i64 [[TMP17]], [[CONV26]]
// CHECK2-NEXT:    [[MUL28:%.*]] = mul nsw i64 [[DIV27]], 5
// CHECK2-NEXT:    [[ADD29:%.*]] = add nsw i64 0, [[MUL28]]
// CHECK2-NEXT:    [[CONV30:%.*]] = trunc i64 [[ADD29]] to i32
// CHECK2-NEXT:    store i32 [[CONV30]], i32* [[DOTFLOOR_0_IV_I11]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CONV31:%.*]] = sext i32 [[TMP20]] to i64
// CHECK2-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB32:%.*]] = sub i32 [[TMP23]], [[TMP24]]
// CHECK2-NEXT:    [[SUB33:%.*]] = sub i32 [[SUB32]], 1
// CHECK2-NEXT:    [[ADD34:%.*]] = add i32 [[SUB33]], 1
// CHECK2-NEXT:    [[DIV35:%.*]] = udiv i32 [[ADD34]], 1
// CHECK2-NEXT:    [[MUL36:%.*]] = mul i32 1, [[DIV35]]
// CHECK2-NEXT:    [[MUL37:%.*]] = mul i32 [[MUL36]], 4
// CHECK2-NEXT:    [[CONV38:%.*]] = zext i32 [[MUL37]] to i64
// CHECK2-NEXT:    [[DIV39:%.*]] = sdiv i64 [[TMP22]], [[CONV38]]
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB40:%.*]] = sub i32 [[TMP25]], [[TMP26]]
// CHECK2-NEXT:    [[SUB41:%.*]] = sub i32 [[SUB40]], 1
// CHECK2-NEXT:    [[ADD42:%.*]] = add i32 [[SUB41]], 1
// CHECK2-NEXT:    [[DIV43:%.*]] = udiv i32 [[ADD42]], 1
// CHECK2-NEXT:    [[MUL44:%.*]] = mul i32 1, [[DIV43]]
// CHECK2-NEXT:    [[MUL45:%.*]] = mul i32 [[MUL44]], 4
// CHECK2-NEXT:    [[CONV46:%.*]] = zext i32 [[MUL45]] to i64
// CHECK2-NEXT:    [[MUL47:%.*]] = mul nsw i64 [[DIV39]], [[CONV46]]
// CHECK2-NEXT:    [[SUB48:%.*]] = sub nsw i64 [[TMP21]], [[MUL47]]
// CHECK2-NEXT:    [[DIV49:%.*]] = sdiv i64 [[SUB48]], 4
// CHECK2-NEXT:    [[MUL50:%.*]] = mul nsw i64 [[DIV49]], 1
// CHECK2-NEXT:    [[ADD51:%.*]] = add nsw i64 [[CONV31]], [[MUL50]]
// CHECK2-NEXT:    [[CONV52:%.*]] = trunc i64 [[ADD51]] to i32
// CHECK2-NEXT:    store i32 [[CONV52]], i32* [[DOTTILE_0_IV_I12]], align 4
// CHECK2-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB53:%.*]] = sub i32 [[TMP29]], [[TMP30]]
// CHECK2-NEXT:    [[SUB54:%.*]] = sub i32 [[SUB53]], 1
// CHECK2-NEXT:    [[ADD55:%.*]] = add i32 [[SUB54]], 1
// CHECK2-NEXT:    [[DIV56:%.*]] = udiv i32 [[ADD55]], 1
// CHECK2-NEXT:    [[MUL57:%.*]] = mul i32 1, [[DIV56]]
// CHECK2-NEXT:    [[MUL58:%.*]] = mul i32 [[MUL57]], 4
// CHECK2-NEXT:    [[CONV59:%.*]] = zext i32 [[MUL58]] to i64
// CHECK2-NEXT:    [[DIV60:%.*]] = sdiv i64 [[TMP28]], [[CONV59]]
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB61:%.*]] = sub i32 [[TMP31]], [[TMP32]]
// CHECK2-NEXT:    [[SUB62:%.*]] = sub i32 [[SUB61]], 1
// CHECK2-NEXT:    [[ADD63:%.*]] = add i32 [[SUB62]], 1
// CHECK2-NEXT:    [[DIV64:%.*]] = udiv i32 [[ADD63]], 1
// CHECK2-NEXT:    [[MUL65:%.*]] = mul i32 1, [[DIV64]]
// CHECK2-NEXT:    [[MUL66:%.*]] = mul i32 [[MUL65]], 4
// CHECK2-NEXT:    [[CONV67:%.*]] = zext i32 [[MUL66]] to i64
// CHECK2-NEXT:    [[MUL68:%.*]] = mul nsw i64 [[DIV60]], [[CONV67]]
// CHECK2-NEXT:    [[SUB69:%.*]] = sub nsw i64 [[TMP27]], [[MUL68]]
// CHECK2-NEXT:    [[TMP33:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB70:%.*]] = sub i32 [[TMP35]], [[TMP36]]
// CHECK2-NEXT:    [[SUB71:%.*]] = sub i32 [[SUB70]], 1
// CHECK2-NEXT:    [[ADD72:%.*]] = add i32 [[SUB71]], 1
// CHECK2-NEXT:    [[DIV73:%.*]] = udiv i32 [[ADD72]], 1
// CHECK2-NEXT:    [[MUL74:%.*]] = mul i32 1, [[DIV73]]
// CHECK2-NEXT:    [[MUL75:%.*]] = mul i32 [[MUL74]], 4
// CHECK2-NEXT:    [[CONV76:%.*]] = zext i32 [[MUL75]] to i64
// CHECK2-NEXT:    [[DIV77:%.*]] = sdiv i64 [[TMP34]], [[CONV76]]
// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB78:%.*]] = sub i32 [[TMP37]], [[TMP38]]
// CHECK2-NEXT:    [[SUB79:%.*]] = sub i32 [[SUB78]], 1
// CHECK2-NEXT:    [[ADD80:%.*]] = add i32 [[SUB79]], 1
// CHECK2-NEXT:    [[DIV81:%.*]] = udiv i32 [[ADD80]], 1
// CHECK2-NEXT:    [[MUL82:%.*]] = mul i32 1, [[DIV81]]
// CHECK2-NEXT:    [[MUL83:%.*]] = mul i32 [[MUL82]], 4
// CHECK2-NEXT:    [[CONV84:%.*]] = zext i32 [[MUL83]] to i64
// CHECK2-NEXT:    [[MUL85:%.*]] = mul nsw i64 [[DIV77]], [[CONV84]]
// CHECK2-NEXT:    [[SUB86:%.*]] = sub nsw i64 [[TMP33]], [[MUL85]]
// CHECK2-NEXT:    [[DIV87:%.*]] = sdiv i64 [[SUB86]], 4
// CHECK2-NEXT:    [[MUL88:%.*]] = mul nsw i64 [[DIV87]], 4
// CHECK2-NEXT:    [[SUB89:%.*]] = sub nsw i64 [[SUB69]], [[MUL88]]
// CHECK2-NEXT:    [[MUL90:%.*]] = mul nsw i64 [[SUB89]], 3
// CHECK2-NEXT:    [[ADD91:%.*]] = add nsw i64 7, [[MUL90]]
// CHECK2-NEXT:    [[CONV92:%.*]] = trunc i64 [[ADD91]] to i32
// CHECK2-NEXT:    store i32 [[CONV92]], i32* [[J13]], align 4
// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTTILE_0_IV_I12]], align 4
// CHECK2-NEXT:    [[MUL93:%.*]] = mul nsw i32 [[TMP39]], 3
// CHECK2-NEXT:    [[ADD94:%.*]] = add nsw i32 7, [[MUL93]]
// CHECK2-NEXT:    store i32 [[ADD94]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[J13]], align 4
// CHECK2-NEXT:    call void (...) @body(i32 [[TMP40]], i32 [[TMP41]])
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP42:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
// CHECK2-NEXT:    [[ADD95:%.*]] = add nsw i64 [[TMP42]], 1
// CHECK2-NEXT:    store i64 [[ADD95]], i64* [[DOTOMP_IV]], align 8
// 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:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@foo6
// CHECK2-SAME: () #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I:%.*]] = 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 7, i32* [[I]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 0, 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:    [[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* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 0
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 0, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK2-NEXT:    [[CMP3:%.*]] = icmp slt i32 4, [[ADD2]]
// CHECK2-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
// CHECK2:       cond.true4:
// CHECK2-NEXT:    br label [[COND_END7:%.*]]
// CHECK2:       cond.false5:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 5
// CHECK2-NEXT:    br label [[COND_END7]]
// CHECK2:       cond.end7:
// CHECK2-NEXT:    [[COND8:%.*]] = phi i32 [ 4, [[COND_TRUE4]] ], [ [[ADD6]], [[COND_FALSE5]] ]
// CHECK2-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP9]], [[COND8]]
// CHECK2-NEXT:    br i1 [[CMP9]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP12]], 3
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 7, [[MUL10]]
// CHECK2-NEXT:    store i32 [[ADD11]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    call void (...) @body(i32 [[TMP13]])
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK2-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@tfoo7
// CHECK2-SAME: () #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @_Z4foo7IiLi3ELi5EEvT_S0_(i32 0, i32 42)
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z4foo7IiLi3ELi5EEvT_S0_
// CHECK2-SAME: (i32 [[START:%.*]], i32 [[END:%.*]]) #[[ATTR2]] comdat {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[START_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[END_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32 [[START]], i32* [[START_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[END]], i32* [[END_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[START_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[START_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[END_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
// CHECK2-NEXT:    [[SUB3:%.*]] = sub i32 [[SUB]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add i32 [[SUB3]], 3
// CHECK2-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 3
// CHECK2-NEXT:    [[SUB4:%.*]] = sub i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB4]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
// CHECK2:       for.cond:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[ADD5:%.*]] = add i32 [[TMP6]], 1
// CHECK2-NEXT:    [[CMP:%.*]] = icmp ult i32 [[TMP5]], [[ADD5]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END17:%.*]]
// CHECK2:       for.body:
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND6:%.*]]
// CHECK2:       for.cond6:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[ADD7:%.*]] = add i32 [[TMP9]], 1
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP10]], 5
// CHECK2-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[ADD7]], [[ADD8]]
// CHECK2-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK2-NEXT:    [[ADD10:%.*]] = add i32 [[TMP11]], 1
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP12]], 5
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[ADD10]], [[COND_TRUE]] ], [ [[ADD11]], [[COND_FALSE]] ]
// CHECK2-NEXT:    [[CMP12:%.*]] = icmp ult i32 [[TMP8]], [[COND]]
// CHECK2-NEXT:    br i1 [[CMP12]], label [[FOR_BODY13:%.*]], label [[FOR_END:%.*]]
// CHECK2:       for.body13:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul i32 [[TMP14]], 3
// CHECK2-NEXT:    [[ADD14:%.*]] = add i32 [[TMP13]], [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    call void (...) @body(i32 [[TMP15]])
// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
// CHECK2:       for.inc:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP16]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTTILE_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND6]], !llvm.loop [[LOOP20:![0-9]+]]
// CHECK2:       for.end:
// CHECK2-NEXT:    br label [[FOR_INC15:%.*]]
// CHECK2:       for.inc15:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP17]], 5
// CHECK2-NEXT:    store i32 [[ADD16]], i32* [[DOTFLOOR_0_IV_I]], align 4
// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
// CHECK2:       for.end17:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_tile_codegen.cpp
// CHECK2-SAME: () #[[ATTR0]] section ".text.startup" {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__cxx_global_var_init()
// CHECK2-NEXT:    ret void
//