// 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 _
// expected-no-diagnostics
#ifndef HEADER
#define HEADER
// Test host codegen.
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK2
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK4
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5

// RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK6
// RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK7
// RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK8
// RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9

// Test target codegen - host bc file has to be created first. (no significant differences with host version of target region)
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK10
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
// RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK12
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK13

// RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK14
// RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK15
// RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
// RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK16
// RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
// RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK17

#ifdef CK1


int target_teams_fun(int *g){
  int n = 1000;
  int a[1000];
  int te = n / 128;
  int th = 128;
  // discard n_addr
  // discard capture expressions for te and th

  #pragma omp target teams distribute parallel for num_teams(te), thread_limit(th)
  for(int i = 0; i < n; i++) {
    a[i] = 0;
    #pragma omp cancel for
  }

  {{{
  #pragma omp target teams distribute parallel for is_device_ptr(g)
  for(int i = 0; i < n; i++) {
    a[i] = g[0];
  }
  }}}

  // outlined target regions


  return a[0];
}

#endif // CK1
#endif // HEADER
// CHECK1-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK1-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK1-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[N_CASTED5:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK1-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK1-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK1-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV2]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED3]] to i32*
// CHECK1-NEXT:    store i32 [[TMP7]], i32* [[CONV4]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED3]], align 8
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51(i64 [[TMP4]], [1000 x i32]* [[A]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2:[0-9]+]]
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[N]], align 4
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED5]] to i32*
// CHECK1-NEXT:    store i32 [[TMP9]], i32* [[CONV6]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[N_CASTED5]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58(i64 [[TMP10]], [1000 x i32]* [[A]], i32* [[TMP11]]) #[[ATTR2]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    ret i32 [[TMP12]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK1-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]], i64 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
// CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV3]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV4]], align 8
// CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// 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 i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK1-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK1:       omp.precond.then:
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK1-NEXT:    store i32 [[ADD]], 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:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       omp.precond.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = 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:    [[I5:%.*]] = 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 i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK1-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK1:       omp.precond.then:
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK1-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV4]], 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:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I5]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK1-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK1-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK1:       .cancel.exit:
// CHECK1-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK1:       .cancel.continue:
// 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:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       cancel.exit:
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK1:       omp.precond.end:
// CHECK1-NEXT:    br label [[CANCEL_CONT]]
// CHECK1:       cancel.cont:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK1-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV1]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// 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 i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK1-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK1:       omp.precond.then:
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]], i32* [[TMP20]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       omp.precond.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = 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:    [[I5:%.*]] = 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 i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK1-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK1-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK1:       omp.precond.then:
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK1-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV4]], 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:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i64 0
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I5]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX8]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK1-NEXT:    store i32 [[ADD9]], 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:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       omp.precond.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK2-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK2-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[N_CASTED9:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [3 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [3 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [3 x i8*], align 8
// CHECK2-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_15:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_16:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK2-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK2-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK2-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[CONV2]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED3]] to i32*
// CHECK2-NEXT:    store i32 [[TMP7]], i32* [[CONV4]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED3]], align 8
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK2-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK2-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [1000 x i32]**
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [1000 x i32]**
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP17]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP20]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP22]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
// CHECK2-NEXT:    store i64 [[TMP8]], i64* [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
// CHECK2-NEXT:    store i64 [[TMP8]], i64* [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK2-NEXT:    store i8* null, i8** [[TMP28]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK2-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP34]], 0
// CHECK2-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[DIV7]], 1
// CHECK2-NEXT:    store i32 [[SUB8]], i32* [[DOTCAPTURE_EXPR_6]], align 4
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_6]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP35]], 1
// CHECK2-NEXT:    [[TMP36:%.*]] = zext i32 [[ADD]] to i64
// CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP36]])
// CHECK2-NEXT:    [[TMP37:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.region_id, i32 4, i8** [[TMP29]], i8** [[TMP30]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP31]], i32 [[TMP32]])
// CHECK2-NEXT:    [[TMP38:%.*]] = icmp ne i32 [[TMP37]], 0
// CHECK2-NEXT:    br i1 [[TMP38]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51(i64 [[TMP4]], [1000 x i32]* [[A]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2:[0-9]+]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[N]], align 4
// CHECK2-NEXT:    [[CONV10:%.*]] = bitcast i64* [[N_CASTED9]] to i32*
// CHECK2-NEXT:    store i32 [[TMP39]], i32* [[CONV10]], align 4
// CHECK2-NEXT:    [[TMP40:%.*]] = load i64, i64* [[N_CASTED9]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK2-NEXT:    store i64 [[TMP40]], i64* [[TMP43]], align 8
// CHECK2-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
// CHECK2-NEXT:    store i64 [[TMP40]], i64* [[TMP45]], align 8
// CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP46]], align 8
// CHECK2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to [1000 x i32]**
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP48]], align 8
// CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to [1000 x i32]**
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP50]], align 8
// CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP51]], align 8
// CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i32**
// CHECK2-NEXT:    store i32* [[TMP41]], i32** [[TMP53]], align 8
// CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to i32**
// CHECK2-NEXT:    store i32* [[TMP41]], i32** [[TMP55]], align 8
// CHECK2-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP56]], align 8
// CHECK2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP59:%.*]] = load i32, i32* [[N]], align 4
// CHECK2-NEXT:    store i32 [[TMP59]], i32* [[DOTCAPTURE_EXPR_15]], align 4
// CHECK2-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_15]], align 4
// CHECK2-NEXT:    [[SUB17:%.*]] = sub nsw i32 [[TMP60]], 0
// CHECK2-NEXT:    [[DIV18:%.*]] = sdiv i32 [[SUB17]], 1
// CHECK2-NEXT:    [[SUB19:%.*]] = sub nsw i32 [[DIV18]], 1
// CHECK2-NEXT:    store i32 [[SUB19]], i32* [[DOTCAPTURE_EXPR_16]], align 4
// CHECK2-NEXT:    [[TMP61:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_16]], align 4
// CHECK2-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP61]], 1
// CHECK2-NEXT:    [[TMP62:%.*]] = zext i32 [[ADD20]] to i64
// CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP62]])
// CHECK2-NEXT:    [[TMP63:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.region_id, i32 3, i8** [[TMP57]], i8** [[TMP58]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK2-NEXT:    [[TMP64:%.*]] = icmp ne i32 [[TMP63]], 0
// CHECK2-NEXT:    br i1 [[TMP64]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK2:       omp_offload.failed21:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58(i64 [[TMP40]], [1000 x i32]* [[A]], i32* [[TMP41]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK2:       omp_offload.cont22:
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP65:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    ret i32 [[TMP65]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK2-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]], i64 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
// CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV3]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV4]], align 8
// CHECK2-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// 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 i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK2-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]])
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK2-NEXT:    store i32 [[ADD]], 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:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = 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:    [[I5:%.*]] = 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 i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK2-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV4]], 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:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I5]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK2-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK2-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK2:       .cancel.exit:
// CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK2:       .cancel.continue:
// 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:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       cancel.exit:
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    br label [[CANCEL_CONT]]
// CHECK2:       cancel.cont:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK2-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV1]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// 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 i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK2-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]], i32* [[TMP20]])
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = 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:    [[I5:%.*]] = 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 i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK2-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK2-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK2:       omp.precond.then:
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK2-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV4]], 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:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i64 0
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I5]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX8]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK2-NEXT:    store i32 [[ADD9]], 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:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK2-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK3-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK3-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[N_CASTED9:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [3 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [3 x i8*], align 8
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [3 x i8*], align 8
// CHECK3-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_15:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_16:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK3-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK3-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK3-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[CONV2]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED3]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[CONV4]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED3]], align 8
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK3-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK3-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [1000 x i32]**
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP15]], align 8
// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [1000 x i32]**
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP17]], align 8
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK3-NEXT:    store i64 [[TMP6]], i64* [[TMP20]], align 8
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK3-NEXT:    store i64 [[TMP6]], i64* [[TMP22]], align 8
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
// CHECK3-NEXT:    store i64 [[TMP8]], i64* [[TMP25]], align 8
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
// CHECK3-NEXT:    store i64 [[TMP8]], i64* [[TMP27]], align 8
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP28]], align 8
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK3-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP34]], 0
// CHECK3-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB]], 1
// CHECK3-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[DIV7]], 1
// CHECK3-NEXT:    store i32 [[SUB8]], i32* [[DOTCAPTURE_EXPR_6]], align 4
// CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_6]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP35]], 1
// CHECK3-NEXT:    [[TMP36:%.*]] = zext i32 [[ADD]] to i64
// CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP36]])
// CHECK3-NEXT:    [[TMP37:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.region_id, i32 4, i8** [[TMP29]], i8** [[TMP30]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP31]], i32 [[TMP32]])
// CHECK3-NEXT:    [[TMP38:%.*]] = icmp ne i32 [[TMP37]], 0
// CHECK3-NEXT:    br i1 [[TMP38]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51(i64 [[TMP4]], [1000 x i32]* [[A]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2:[0-9]+]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[N]], align 4
// CHECK3-NEXT:    [[CONV10:%.*]] = bitcast i64* [[N_CASTED9]] to i32*
// CHECK3-NEXT:    store i32 [[TMP39]], i32* [[CONV10]], align 4
// CHECK3-NEXT:    [[TMP40:%.*]] = load i64, i64* [[N_CASTED9]], align 8
// CHECK3-NEXT:    [[TMP41:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK3-NEXT:    store i64 [[TMP40]], i64* [[TMP43]], align 8
// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
// CHECK3-NEXT:    store i64 [[TMP40]], i64* [[TMP45]], align 8
// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP46]], align 8
// CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to [1000 x i32]**
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP48]], align 8
// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to [1000 x i32]**
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP50]], align 8
// CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP51]], align 8
// CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i32**
// CHECK3-NEXT:    store i32* [[TMP41]], i32** [[TMP53]], align 8
// CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to i32**
// CHECK3-NEXT:    store i32* [[TMP41]], i32** [[TMP55]], align 8
// CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP56]], align 8
// CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP59:%.*]] = load i32, i32* [[N]], align 4
// CHECK3-NEXT:    store i32 [[TMP59]], i32* [[DOTCAPTURE_EXPR_15]], align 4
// CHECK3-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_15]], align 4
// CHECK3-NEXT:    [[SUB17:%.*]] = sub nsw i32 [[TMP60]], 0
// CHECK3-NEXT:    [[DIV18:%.*]] = sdiv i32 [[SUB17]], 1
// CHECK3-NEXT:    [[SUB19:%.*]] = sub nsw i32 [[DIV18]], 1
// CHECK3-NEXT:    store i32 [[SUB19]], i32* [[DOTCAPTURE_EXPR_16]], align 4
// CHECK3-NEXT:    [[TMP61:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_16]], align 4
// CHECK3-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP61]], 1
// CHECK3-NEXT:    [[TMP62:%.*]] = zext i32 [[ADD20]] to i64
// CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP62]])
// CHECK3-NEXT:    [[TMP63:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.region_id, i32 3, i8** [[TMP57]], i8** [[TMP58]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK3-NEXT:    [[TMP64:%.*]] = icmp ne i32 [[TMP63]], 0
// CHECK3-NEXT:    br i1 [[TMP64]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK3:       omp_offload.failed21:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58(i64 [[TMP40]], [1000 x i32]* [[A]], i32* [[TMP41]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK3:       omp_offload.cont22:
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 0
// CHECK3-NEXT:    [[TMP65:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    ret i32 [[TMP65]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK3-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]], i64 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
// CHECK3-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK3-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV3]], align 8
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV4]], align 8
// CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK3-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK3:       omp.precond.then:
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK3-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK3-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK3:       omp.precond.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I5:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK3-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK3:       omp.precond.then:
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK3-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK3-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK3-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I5]], align 4
// CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK3-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK3-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK3-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK3:       .cancel.exit:
// CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK3:       .cancel.continue:
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK3:       cancel.exit:
// CHECK3-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK3:       omp.precond.end:
// CHECK3-NEXT:    br label [[CANCEL_CONT]]
// CHECK3:       cancel.cont:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK3-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK3-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK3:       omp.precond.then:
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK3-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK3-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK3-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]], i32* [[TMP20]])
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK3:       omp.precond.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I5:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK3-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK3-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK3:       omp.precond.then:
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK3-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK3-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK3-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK3-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i64 0
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I5]], align 4
// CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK3-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK3-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX8]], align 4
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK3-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK3:       omp.precond.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK4-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK4-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
// CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[N_CASTED7:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [3 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [3 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [3 x i8*], align 4
// CHECK4-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK4-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK4-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK4-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [1000 x i32]**
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP15]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [1000 x i32]**
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP17]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK4-NEXT:    store i32 [[TMP6]], i32* [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK4-NEXT:    store i32 [[TMP6]], i32* [[TMP22]], align 4
// CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
// CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP25]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
// CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP27]], align 4
// CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK4-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK4-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK4-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP34]], 0
// CHECK4-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB]], 1
// CHECK4-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[DIV5]], 1
// CHECK4-NEXT:    store i32 [[SUB6]], i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK4-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP35]], 1
// CHECK4-NEXT:    [[TMP36:%.*]] = zext i32 [[ADD]] to i64
// CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP36]])
// CHECK4-NEXT:    [[TMP37:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.region_id, i32 4, i8** [[TMP29]], i8** [[TMP30]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP31]], i32 [[TMP32]])
// CHECK4-NEXT:    [[TMP38:%.*]] = icmp ne i32 [[TMP37]], 0
// CHECK4-NEXT:    br i1 [[TMP38]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51(i32 [[TMP4]], [1000 x i32]* [[A]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2:[0-9]+]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    [[TMP39:%.*]] = load i32, i32* [[N]], align 4
// CHECK4-NEXT:    store i32 [[TMP39]], i32* [[N_CASTED7]], align 4
// CHECK4-NEXT:    [[TMP40:%.*]] = load i32, i32* [[N_CASTED7]], align 4
// CHECK4-NEXT:    [[TMP41:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i32*
// CHECK4-NEXT:    store i32 [[TMP40]], i32* [[TMP43]], align 4
// CHECK4-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i32*
// CHECK4-NEXT:    store i32 [[TMP40]], i32* [[TMP45]], align 4
// CHECK4-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP46]], align 4
// CHECK4-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to [1000 x i32]**
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP48]], align 4
// CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to [1000 x i32]**
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP50]], align 4
// CHECK4-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP51]], align 4
// CHECK4-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i32**
// CHECK4-NEXT:    store i32* [[TMP41]], i32** [[TMP53]], align 4
// CHECK4-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to i32**
// CHECK4-NEXT:    store i32* [[TMP41]], i32** [[TMP55]], align 4
// CHECK4-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP56]], align 4
// CHECK4-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP59:%.*]] = load i32, i32* [[N]], align 4
// CHECK4-NEXT:    store i32 [[TMP59]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK4-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK4-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP60]], 0
// CHECK4-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
// CHECK4-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
// CHECK4-NEXT:    store i32 [[SUB16]], i32* [[DOTCAPTURE_EXPR_13]], align 4
// CHECK4-NEXT:    [[TMP61:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_13]], align 4
// CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP61]], 1
// CHECK4-NEXT:    [[TMP62:%.*]] = zext i32 [[ADD17]] to i64
// CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP62]])
// CHECK4-NEXT:    [[TMP63:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.region_id, i32 3, i8** [[TMP57]], i8** [[TMP58]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK4-NEXT:    [[TMP64:%.*]] = icmp ne i32 [[TMP63]], 0
// CHECK4-NEXT:    br i1 [[TMP64]], label [[OMP_OFFLOAD_FAILED18:%.*]], label [[OMP_OFFLOAD_CONT19:%.*]]
// CHECK4:       omp_offload.failed18:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58(i32 [[TMP40]], [1000 x i32]* [[A]], i32* [[TMP41]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT19]]
// CHECK4:       omp_offload.cont19:
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP65:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    ret i32 [[TMP65]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK4-SAME: (i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]], i32 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK4-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK4-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK4-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK4:       omp.precond.then:
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK4-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4:       cond.true:
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK4-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], [1000 x i32]* [[TMP0]])
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK4:       omp.loop.exit:
// CHECK4-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK4:       omp.precond.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK4-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK4-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK4:       omp.precond.then:
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK4-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4:       cond.true:
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK4-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP17]]
// CHECK4-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK4-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK4-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK4:       .cancel.exit:
// CHECK4-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK4:       .cancel.continue:
// CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK4:       omp.body.continue:
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK4-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK4:       omp.loop.exit:
// CHECK4-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK4:       cancel.exit:
// CHECK4-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK4:       omp.precond.end:
// CHECK4-NEXT:    br label [[CANCEL_CONT]]
// CHECK4:       cancel.cont:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK4-SAME: (i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK4-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK4-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK4:       omp.precond.then:
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK4-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4:       cond.true:
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK4-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], [1000 x i32]* [[TMP0]], i32* [[TMP18]])
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK4:       omp.loop.exit:
// CHECK4-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK4:       omp.precond.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK4-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK4-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK4:       omp.precond.then:
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK4-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4:       cond.true:
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK4-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i32 0
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
// CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP19]]
// CHECK4-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX6]], align 4
// CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK4:       omp.body.continue:
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK4-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK4:       omp.loop.exit:
// CHECK4-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK4:       omp.precond.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK4-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK5-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK5-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[N_CASTED7:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [3 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [3 x i8*], align 4
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [3 x i8*], align 4
// CHECK5-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK5-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK5-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK5-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK5-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N]], align 4
// CHECK5-NEXT:    store i32 [[TMP3]], i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK5-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK5-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK5-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [1000 x i32]**
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP15]], align 4
// CHECK5-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK5-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [1000 x i32]**
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP17]], align 4
// CHECK5-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK5-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK5-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK5-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK5-NEXT:    store i32 [[TMP6]], i32* [[TMP20]], align 4
// CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK5-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK5-NEXT:    store i32 [[TMP6]], i32* [[TMP22]], align 4
// CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK5-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK5-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
// CHECK5-NEXT:    store i32 [[TMP8]], i32* [[TMP25]], align 4
// CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK5-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
// CHECK5-NEXT:    store i32 [[TMP8]], i32* [[TMP27]], align 4
// CHECK5-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK5-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK5-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    [[TMP33:%.*]] = load i32, i32* [[N]], align 4
// CHECK5-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK5-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP34]], 0
// CHECK5-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB]], 1
// CHECK5-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[DIV5]], 1
// CHECK5-NEXT:    store i32 [[SUB6]], i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK5-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP35]], 1
// CHECK5-NEXT:    [[TMP36:%.*]] = zext i32 [[ADD]] to i64
// CHECK5-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP36]])
// CHECK5-NEXT:    [[TMP37:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.region_id, i32 4, i8** [[TMP29]], i8** [[TMP30]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP31]], i32 [[TMP32]])
// CHECK5-NEXT:    [[TMP38:%.*]] = icmp ne i32 [[TMP37]], 0
// CHECK5-NEXT:    br i1 [[TMP38]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK5:       omp_offload.failed:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51(i32 [[TMP4]], [1000 x i32]* [[A]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2:[0-9]+]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK5:       omp_offload.cont:
// CHECK5-NEXT:    [[TMP39:%.*]] = load i32, i32* [[N]], align 4
// CHECK5-NEXT:    store i32 [[TMP39]], i32* [[N_CASTED7]], align 4
// CHECK5-NEXT:    [[TMP40:%.*]] = load i32, i32* [[N_CASTED7]], align 4
// CHECK5-NEXT:    [[TMP41:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK5-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i32*
// CHECK5-NEXT:    store i32 [[TMP40]], i32* [[TMP43]], align 4
// CHECK5-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i32*
// CHECK5-NEXT:    store i32 [[TMP40]], i32* [[TMP45]], align 4
// CHECK5-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
// CHECK5-NEXT:    store i8* null, i8** [[TMP46]], align 4
// CHECK5-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK5-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to [1000 x i32]**
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP48]], align 4
// CHECK5-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK5-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to [1000 x i32]**
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[TMP50]], align 4
// CHECK5-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
// CHECK5-NEXT:    store i8* null, i8** [[TMP51]], align 4
// CHECK5-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 2
// CHECK5-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i32**
// CHECK5-NEXT:    store i32* [[TMP41]], i32** [[TMP53]], align 4
// CHECK5-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 2
// CHECK5-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to i32**
// CHECK5-NEXT:    store i32* [[TMP41]], i32** [[TMP55]], align 4
// CHECK5-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 2
// CHECK5-NEXT:    store i8* null, i8** [[TMP56]], align 4
// CHECK5-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP59:%.*]] = load i32, i32* [[N]], align 4
// CHECK5-NEXT:    store i32 [[TMP59]], i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK5-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
// CHECK5-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP60]], 0
// CHECK5-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
// CHECK5-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
// CHECK5-NEXT:    store i32 [[SUB16]], i32* [[DOTCAPTURE_EXPR_13]], align 4
// CHECK5-NEXT:    [[TMP61:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_13]], align 4
// CHECK5-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP61]], 1
// CHECK5-NEXT:    [[TMP62:%.*]] = zext i32 [[ADD17]] to i64
// CHECK5-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP62]])
// CHECK5-NEXT:    [[TMP63:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.region_id, i32 3, i8** [[TMP57]], i8** [[TMP58]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
// CHECK5-NEXT:    [[TMP64:%.*]] = icmp ne i32 [[TMP63]], 0
// CHECK5-NEXT:    br i1 [[TMP64]], label [[OMP_OFFLOAD_FAILED18:%.*]], label [[OMP_OFFLOAD_CONT19:%.*]]
// CHECK5:       omp_offload.failed18:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58(i32 [[TMP40]], [1000 x i32]* [[A]], i32* [[TMP41]]) #[[ATTR2]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT19]]
// CHECK5:       omp_offload.cont19:
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP65:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK5-NEXT:    ret i32 [[TMP65]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK5-SAME: (i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]], i32 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
// CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK5-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP4]], i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK5-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK5-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK5-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK5:       omp.precond.then:
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK5-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], [1000 x i32]* [[TMP0]])
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK5-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK5:       omp.precond.end:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK5-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK5-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK5-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK5:       omp.precond.then:
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK5-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP17]]
// CHECK5-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK5-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK5-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK5-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK5-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK5:       .cancel.exit:
// CHECK5-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK5:       .cancel.continue:
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK5-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK5-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK5:       cancel.exit:
// CHECK5-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK5-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK5:       omp.precond.end:
// CHECK5-NEXT:    br label [[CANCEL_CONT]]
// CHECK5:       cancel.cont:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK5-SAME: (i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK5-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK5-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK5-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK5-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK5-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK5:       omp.precond.then:
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK5-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK5-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], [1000 x i32]* [[TMP0]], i32* [[TMP18]])
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK5-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK5:       omp.precond.end:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR1]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK5-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK5-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK5-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK5-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK5:       omp.precond.then:
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK5-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK5-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK5-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i32 0
// CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
// CHECK5-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP19]]
// CHECK5-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX6]], align 4
// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK5:       omp.body.continue:
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK5-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK5-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK5-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK5:       omp.precond.end:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK5-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK5-NEXT:    ret void
//
//
// CHECK6-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK6-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK6-NEXT:  entry:
// CHECK6-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK6-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK6-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    [[I2:%.*]] = alloca i32, align 4
// CHECK6-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK6-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK6-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK6-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK6-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK6-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK6-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK6-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK6-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK6-NEXT:    br label [[FOR_COND:%.*]]
// CHECK6:       for.cond:
// CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[I]], align 4
// CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK6-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP3]], [[TMP4]]
// CHECK6-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK6:       for.body:
// CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[I]], align 4
// CHECK6-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP5]] to i64
// CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK6-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK6-NEXT:    br label [[FOR_INC:%.*]]
// CHECK6:       for.inc:
// CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
// CHECK6-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK6-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK6-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK6:       for.end:
// CHECK6-NEXT:    store i32 0, i32* [[I2]], align 4
// CHECK6-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK6:       for.cond3:
// CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I2]], align 4
// CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N]], align 4
// CHECK6-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK6-NEXT:    br i1 [[CMP4]], label [[FOR_BODY5:%.*]], label [[FOR_END11:%.*]]
// CHECK6:       for.body5:
// CHECK6-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK6-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i64 0
// CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX6]], align 4
// CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I2]], align 4
// CHECK6-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP11]] to i64
// CHECK6-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 [[IDXPROM7]]
// CHECK6-NEXT:    store i32 [[TMP10]], i32* [[ARRAYIDX8]], align 4
// CHECK6-NEXT:    br label [[FOR_INC9:%.*]]
// CHECK6:       for.inc9:
// CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I2]], align 4
// CHECK6-NEXT:    [[INC10:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK6-NEXT:    store i32 [[INC10]], i32* [[I2]], align 4
// CHECK6-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK6:       for.end11:
// CHECK6-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 0
// CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX12]], align 4
// CHECK6-NEXT:    ret i32 [[TMP13]]
//
//
// CHECK7-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK7-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK7-NEXT:  entry:
// CHECK7-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK7-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK7-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    [[I2:%.*]] = alloca i32, align 4
// CHECK7-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK7-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK7-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK7-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK7-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK7-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK7-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK7-NEXT:    br label [[FOR_COND:%.*]]
// CHECK7:       for.cond:
// CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[I]], align 4
// CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK7-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP3]], [[TMP4]]
// CHECK7-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK7:       for.body:
// CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[I]], align 4
// CHECK7-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP5]] to i64
// CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK7-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK7-NEXT:    br label [[FOR_INC:%.*]]
// CHECK7:       for.inc:
// CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
// CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK7-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK7-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK7:       for.end:
// CHECK7-NEXT:    store i32 0, i32* [[I2]], align 4
// CHECK7-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK7:       for.cond3:
// CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I2]], align 4
// CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N]], align 4
// CHECK7-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK7-NEXT:    br i1 [[CMP4]], label [[FOR_BODY5:%.*]], label [[FOR_END11:%.*]]
// CHECK7:       for.body5:
// CHECK7-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK7-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i64 0
// CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX6]], align 4
// CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I2]], align 4
// CHECK7-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP11]] to i64
// CHECK7-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 [[IDXPROM7]]
// CHECK7-NEXT:    store i32 [[TMP10]], i32* [[ARRAYIDX8]], align 4
// CHECK7-NEXT:    br label [[FOR_INC9:%.*]]
// CHECK7:       for.inc9:
// CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I2]], align 4
// CHECK7-NEXT:    [[INC10:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK7-NEXT:    store i32 [[INC10]], i32* [[I2]], align 4
// CHECK7-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK7:       for.end11:
// CHECK7-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 0
// CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX12]], align 4
// CHECK7-NEXT:    ret i32 [[TMP13]]
//
//
// CHECK8-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK8-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK8-NEXT:  entry:
// CHECK8-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK8-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK8-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    [[I2:%.*]] = alloca i32, align 4
// CHECK8-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK8-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK8-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK8-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK8-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK8-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK8-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK8-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK8-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK8-NEXT:    br label [[FOR_COND:%.*]]
// CHECK8:       for.cond:
// CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[I]], align 4
// CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK8-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP3]], [[TMP4]]
// CHECK8-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK8:       for.body:
// CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[I]], align 4
// CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 [[TMP5]]
// CHECK8-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK8-NEXT:    br label [[FOR_INC:%.*]]
// CHECK8:       for.inc:
// CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
// CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK8-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK8-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK8:       for.end:
// CHECK8-NEXT:    store i32 0, i32* [[I2]], align 4
// CHECK8-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK8:       for.cond3:
// CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I2]], align 4
// CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N]], align 4
// CHECK8-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK8-NEXT:    br i1 [[CMP4]], label [[FOR_BODY5:%.*]], label [[FOR_END10:%.*]]
// CHECK8:       for.body5:
// CHECK8-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK8-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i32 0
// CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX6]], align 4
// CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I2]], align 4
// CHECK8-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK8-NEXT:    store i32 [[TMP10]], i32* [[ARRAYIDX7]], align 4
// CHECK8-NEXT:    br label [[FOR_INC8:%.*]]
// CHECK8:       for.inc8:
// CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I2]], align 4
// CHECK8-NEXT:    [[INC9:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK8-NEXT:    store i32 [[INC9]], i32* [[I2]], align 4
// CHECK8-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP5:![0-9]+]]
// CHECK8:       for.end10:
// CHECK8-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 0
// CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX11]], align 4
// CHECK8-NEXT:    ret i32 [[TMP13]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK9-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK9-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK9-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I2:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK9-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK9-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK9-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK9-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK9-NEXT:    br label [[FOR_COND:%.*]]
// CHECK9:       for.cond:
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP3]], [[TMP4]]
// CHECK9-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK9:       for.body:
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[I]], align 4
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 [[TMP5]]
// CHECK9-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    br label [[FOR_INC:%.*]]
// CHECK9:       for.inc:
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
// CHECK9-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK9-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK9-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK9:       for.end:
// CHECK9-NEXT:    store i32 0, i32* [[I2]], align 4
// CHECK9-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK9:       for.cond3:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I2]], align 4
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N]], align 4
// CHECK9-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP4]], label [[FOR_BODY5:%.*]], label [[FOR_END10:%.*]]
// CHECK9:       for.body5:
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i32 0
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX6]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I2]], align 4
// CHECK9-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK9-NEXT:    store i32 [[TMP10]], i32* [[ARRAYIDX7]], align 4
// CHECK9-NEXT:    br label [[FOR_INC8:%.*]]
// CHECK9:       for.inc8:
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I2]], align 4
// CHECK9-NEXT:    [[INC9:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK9-NEXT:    store i32 [[INC9]], i32* [[I2]], align 4
// CHECK9-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP5:![0-9]+]]
// CHECK9:       for.end10:
// CHECK9-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX11]], align 4
// CHECK9-NEXT:    ret i32 [[TMP13]]
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK10-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]], i64 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
// CHECK10-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK10-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK10-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV3]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV4]], align 8
// CHECK10-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK10-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK10-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK10-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK10:       omp.precond.then:
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK10-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    br label [[COND_END]]
// CHECK10:       cond.end:
// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK10-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]])
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK10:       omp.loop.exit:
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK10:       omp.precond.end:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I5:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK10-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK10-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK10-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK10:       omp.precond.then:
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK10-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK10-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK10-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK10-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    br label [[COND_END]]
// CHECK10:       cond.end:
// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK10-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I5]], align 4
// CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK10-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK10-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK10-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK10-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK10:       .cancel.exit:
// CHECK10-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK10:       .cancel.continue:
// CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK10:       omp.body.continue:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK10-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK10:       omp.loop.exit:
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK10:       cancel.exit:
// CHECK10-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK10:       omp.precond.end:
// CHECK10-NEXT:    br label [[CANCEL_CONT]]
// CHECK10:       cancel.cont:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK10-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK10-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK10-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP1]], i32* [[CONV1]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK10-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK10-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK10-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK10-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK10:       omp.precond.then:
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK10-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    br label [[COND_END]]
// CHECK10:       cond.end:
// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK10-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]], i32* [[TMP20]])
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK10:       omp.loop.exit:
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK10:       omp.precond.end:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK10-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    [[I5:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK10-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK10-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK10-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK10:       omp.precond.then:
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK10-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK10-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK10-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK10-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    br label [[COND_END]]
// CHECK10:       cond.end:
// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK10-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i64 0
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I5]], align 4
// CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK10-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX8]], align 4
// CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK10:       omp.body.continue:
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK10-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK10:       omp.inner.for.end:
// CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK10:       omp.loop.exit:
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK10:       omp.precond.end:
// CHECK10-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK11-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]], i64 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
// CHECK11-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK11-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK11-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK11-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV3]], align 8
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV4]], align 8
// CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK11-NEXT:    [[CONV5:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK11-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK11-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK11:       omp.precond.then:
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK11-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK11-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]])
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK11:       omp.precond.end:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I5:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK11-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK11-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK11:       omp.precond.then:
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK11-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK11-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK11-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK11-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK11-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK11-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I5]], align 4
// CHECK11-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK11-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK11-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK11-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK11-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK11:       .cancel.exit:
// CHECK11-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK11:       .cancel.continue:
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK11-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK11:       cancel.exit:
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK11-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK11:       omp.precond.end:
// CHECK11-NEXT:    br label [[CANCEL_CONT]]
// CHECK11:       cancel.cont:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK11-SAME: (i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK11-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK11-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[CONV1]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK11-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK11-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK11-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK11:       omp.precond.then:
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK11-NEXT:    [[CONV6:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK11-NEXT:    store i32 [[TMP18]], i32* [[CONV6]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], [1000 x i32]* [[TMP0]], i32* [[TMP20]])
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK11:       omp.precond.end:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i64 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 8
// CHECK11-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I5:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK11-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK11-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK11-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK11:       omp.precond.then:
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK11-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK11-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK11-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK11-NEXT:    store i32 [[CONV3]], i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[CONV4]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK11-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK11-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I5]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i64 0
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I5]], align 4
// CHECK11-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK11-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK11-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX8]], align 4
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK11-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK11:       omp.precond.end:
// CHECK11-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK12-SAME: (i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]], i32 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
// CHECK12-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK12-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], i32* [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK12-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK12-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK12-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK12-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK12:       omp.precond.then:
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK12-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK12:       cond.true:
// CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    br label [[COND_END:%.*]]
// CHECK12:       cond.false:
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    br label [[COND_END]]
// CHECK12:       cond.end:
// CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK12-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK12:       omp.inner.for.body:
// CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], [1000 x i32]* [[TMP0]])
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK12:       omp.inner.for.inc:
// CHECK12-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK12:       omp.inner.for.end:
// CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK12:       omp.loop.exit:
// CHECK12-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK12:       omp.precond.end:
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK12-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK12-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK12-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK12-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK12:       omp.precond.then:
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK12-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK12-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK12:       cond.true:
// CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    br label [[COND_END:%.*]]
// CHECK12:       cond.false:
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    br label [[COND_END]]
// CHECK12:       cond.end:
// CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK12-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK12:       omp.inner.for.body:
// CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK12-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP17]]
// CHECK12-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK12-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK12-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK12-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK12-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK12:       .cancel.exit:
// CHECK12-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK12:       .cancel.continue:
// CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK12:       omp.body.continue:
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK12:       omp.inner.for.inc:
// CHECK12-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK12-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK12:       omp.inner.for.end:
// CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK12:       omp.loop.exit:
// CHECK12-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK12:       cancel.exit:
// CHECK12-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK12-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK12:       omp.precond.end:
// CHECK12-NEXT:    br label [[CANCEL_CONT]]
// CHECK12:       cancel.cont:
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK12-SAME: (i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK12-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK12-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK12-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK12-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK12-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK12-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK12:       omp.precond.then:
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK12-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK12:       cond.true:
// CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    br label [[COND_END:%.*]]
// CHECK12:       cond.false:
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    br label [[COND_END]]
// CHECK12:       cond.end:
// CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK12-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK12:       omp.inner.for.body:
// CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], [1000 x i32]* [[TMP0]], i32* [[TMP18]])
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK12:       omp.inner.for.inc:
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK12:       omp.inner.for.end:
// CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK12:       omp.loop.exit:
// CHECK12-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK12:       omp.precond.end:
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK12-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK12-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK12-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK12-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK12-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK12:       omp.precond.then:
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK12-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK12-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK12:       cond.true:
// CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    br label [[COND_END:%.*]]
// CHECK12:       cond.false:
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    br label [[COND_END]]
// CHECK12:       cond.end:
// CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK12-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK12:       omp.inner.for.body:
// CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK12-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i32 0
// CHECK12-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
// CHECK12-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP19]]
// CHECK12-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX6]], align 4
// CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK12:       omp.body.continue:
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK12:       omp.inner.for.inc:
// CHECK12-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK12-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK12:       omp.inner.for.end:
// CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK12:       omp.loop.exit:
// CHECK12-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK12:       omp.precond.end:
// CHECK12-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK13-SAME: (i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]], i32 [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
// CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK13-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], i32* [[N_CASTED]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [1000 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP5]], [1000 x i32]* [[TMP1]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK13-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], [1000 x i32]* [[TMP0]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK13-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP17]]
// CHECK13-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
// CHECK13-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP19]], i32 2)
// CHECK13-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK13-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK13:       .cancel.exit:
// CHECK13-NEXT:    br label [[CANCEL_EXIT:%.*]]
// CHECK13:       .cancel.continue:
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK13-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       cancel.exit:
// CHECK13-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK13-NEXT:    br label [[CANCEL_CONT:%.*]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    br label [[CANCEL_CONT]]
// CHECK13:       cancel.cont:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58
// CHECK13-SAME: (i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK13-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [1000 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP2]], [1000 x i32]* [[TMP0]], i32* [[TMP3]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK13-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
// CHECK13-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP16]], i32* [[N_CASTED]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK13-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [1000 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], [1000 x i32]* [[TMP0]], i32* [[TMP18]])
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32 [[N:%.*]], [1000 x i32]* nonnull align 4 dereferenceable(4000) [[A:%.*]], i32* [[G:%.*]]) #[[ATTR0]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca [1000 x i32]*, align 4
// CHECK13-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I3:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store [1000 x i32]* [[A]], [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load [1000 x i32]*, [1000 x i32]** [[A_ADDR]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
// CHECK13-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
// CHECK13-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
// CHECK13-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
// CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK13:       omp.precond.then:
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK13-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], i32* [[I3]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP17]], i32 0
// CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
// CHECK13-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[TMP0]], i32 0, i32 [[TMP19]]
// CHECK13-NEXT:    store i32 [[TMP18]], i32* [[ARRAYIDX6]], align 4
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK13-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
// CHECK13-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK13:       omp.precond.end:
// CHECK13-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK14-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK14-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK14-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    [[I2:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK14-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK14-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK14-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK14-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK14-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK14-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK14-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK14-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK14-NEXT:    br label [[FOR_COND:%.*]]
// CHECK14:       for.cond:
// CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[I]], align 4
// CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP3]], [[TMP4]]
// CHECK14-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK14:       for.body:
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[I]], align 4
// CHECK14-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP5]] to i64
// CHECK14-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK14-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK14-NEXT:    br label [[FOR_INC:%.*]]
// CHECK14:       for.inc:
// CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
// CHECK14-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK14-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK14-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK14:       for.end:
// CHECK14-NEXT:    store i32 0, i32* [[I2]], align 4
// CHECK14-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK14:       for.cond3:
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I2]], align 4
// CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N]], align 4
// CHECK14-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK14-NEXT:    br i1 [[CMP4]], label [[FOR_BODY5:%.*]], label [[FOR_END11:%.*]]
// CHECK14:       for.body5:
// CHECK14-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK14-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i64 0
// CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX6]], align 4
// CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I2]], align 4
// CHECK14-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP11]] to i64
// CHECK14-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 [[IDXPROM7]]
// CHECK14-NEXT:    store i32 [[TMP10]], i32* [[ARRAYIDX8]], align 4
// CHECK14-NEXT:    br label [[FOR_INC9:%.*]]
// CHECK14:       for.inc9:
// CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I2]], align 4
// CHECK14-NEXT:    [[INC10:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK14-NEXT:    store i32 [[INC10]], i32* [[I2]], align 4
// CHECK14-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK14:       for.end11:
// CHECK14-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 0
// CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX12]], align 4
// CHECK14-NEXT:    ret i32 [[TMP13]]
//
//
// CHECK15-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK15-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 8
// CHECK15-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK15-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    [[I2:%.*]] = alloca i32, align 4
// CHECK15-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 8
// CHECK15-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK15-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK15-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK15-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK15-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK15-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK15-NEXT:    br label [[FOR_COND:%.*]]
// CHECK15:       for.cond:
// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[I]], align 4
// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP3]], [[TMP4]]
// CHECK15-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK15:       for.body:
// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[I]], align 4
// CHECK15-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP5]] to i64
// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
// CHECK15-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK15-NEXT:    br label [[FOR_INC:%.*]]
// CHECK15:       for.inc:
// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
// CHECK15-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK15-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK15-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK15:       for.end:
// CHECK15-NEXT:    store i32 0, i32* [[I2]], align 4
// CHECK15-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK15:       for.cond3:
// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I2]], align 4
// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N]], align 4
// CHECK15-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK15-NEXT:    br i1 [[CMP4]], label [[FOR_BODY5:%.*]], label [[FOR_END11:%.*]]
// CHECK15:       for.body5:
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[G_ADDR]], align 8
// CHECK15-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i64 0
// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX6]], align 4
// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I2]], align 4
// CHECK15-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP11]] to i64
// CHECK15-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 [[IDXPROM7]]
// CHECK15-NEXT:    store i32 [[TMP10]], i32* [[ARRAYIDX8]], align 4
// CHECK15-NEXT:    br label [[FOR_INC9:%.*]]
// CHECK15:       for.inc9:
// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I2]], align 4
// CHECK15-NEXT:    [[INC10:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK15-NEXT:    store i32 [[INC10]], i32* [[I2]], align 4
// CHECK15-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP4:![0-9]+]]
// CHECK15:       for.end11:
// CHECK15-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i64 0, i64 0
// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX12]], align 4
// CHECK15-NEXT:    ret i32 [[TMP13]]
//
//
// CHECK16-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK16-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK16-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK16-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    [[I2:%.*]] = alloca i32, align 4
// CHECK16-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK16-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK16-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK16-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK16-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK16-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK16-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK16-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK16-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK16-NEXT:    br label [[FOR_COND:%.*]]
// CHECK16:       for.cond:
// CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[I]], align 4
// CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP3]], [[TMP4]]
// CHECK16-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK16:       for.body:
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[I]], align 4
// CHECK16-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 [[TMP5]]
// CHECK16-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK16-NEXT:    br label [[FOR_INC:%.*]]
// CHECK16:       for.inc:
// CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
// CHECK16-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK16-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK16-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK16:       for.end:
// CHECK16-NEXT:    store i32 0, i32* [[I2]], align 4
// CHECK16-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK16:       for.cond3:
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I2]], align 4
// CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N]], align 4
// CHECK16-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK16-NEXT:    br i1 [[CMP4]], label [[FOR_BODY5:%.*]], label [[FOR_END10:%.*]]
// CHECK16:       for.body5:
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK16-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i32 0
// CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX6]], align 4
// CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I2]], align 4
// CHECK16-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK16-NEXT:    store i32 [[TMP10]], i32* [[ARRAYIDX7]], align 4
// CHECK16-NEXT:    br label [[FOR_INC8:%.*]]
// CHECK16:       for.inc8:
// CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I2]], align 4
// CHECK16-NEXT:    [[INC9:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK16-NEXT:    store i32 [[INC9]], i32* [[I2]], align 4
// CHECK16-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP5:![0-9]+]]
// CHECK16:       for.end10:
// CHECK16-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 0
// CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX11]], align 4
// CHECK16-NEXT:    ret i32 [[TMP13]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK17-SAME: (i32* [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[G_ADDR:%.*]] = alloca i32*, align 4
// CHECK17-NEXT:    [[N:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
// CHECK17-NEXT:    [[TE:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[TH:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[I2:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    store i32* [[G]], i32** [[G_ADDR]], align 4
// CHECK17-NEXT:    store i32 1000, i32* [[N]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK17-NEXT:    store i32 [[DIV]], i32* [[TE]], align 4
// CHECK17-NEXT:    store i32 128, i32* [[TH]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[I]], align 4
// CHECK17-NEXT:    br label [[FOR_COND:%.*]]
// CHECK17:       for.cond:
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[I]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP3]], [[TMP4]]
// CHECK17-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
// CHECK17:       for.body:
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[I]], align 4
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 [[TMP5]]
// CHECK17-NEXT:    store i32 0, i32* [[ARRAYIDX]], align 4
// CHECK17-NEXT:    br label [[FOR_INC:%.*]]
// CHECK17:       for.inc:
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
// CHECK17-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP6]], 1
// CHECK17-NEXT:    store i32 [[INC]], i32* [[I]], align 4
// CHECK17-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK17:       for.end:
// CHECK17-NEXT:    store i32 0, i32* [[I2]], align 4
// CHECK17-NEXT:    br label [[FOR_COND3:%.*]]
// CHECK17:       for.cond3:
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I2]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N]], align 4
// CHECK17-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP7]], [[TMP8]]
// CHECK17-NEXT:    br i1 [[CMP4]], label [[FOR_BODY5:%.*]], label [[FOR_END10:%.*]]
// CHECK17:       for.body5:
// CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[G_ADDR]], align 4
// CHECK17-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i32 0
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[ARRAYIDX6]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I2]], align 4
// CHECK17-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 [[TMP11]]
// CHECK17-NEXT:    store i32 [[TMP10]], i32* [[ARRAYIDX7]], align 4
// CHECK17-NEXT:    br label [[FOR_INC8:%.*]]
// CHECK17:       for.inc8:
// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I2]], align 4
// CHECK17-NEXT:    [[INC9:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK17-NEXT:    store i32 [[INC9]], i32* [[I2]], align 4
// CHECK17-NEXT:    br label [[FOR_COND3]], !llvm.loop [[LOOP5:![0-9]+]]
// CHECK17:       for.end10:
// CHECK17-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [1000 x i32], [1000 x i32]* [[A]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX11]], align 4
// CHECK17-NEXT:    ret i32 [[TMP13]]
//