test/OpenMP/target_teams_distribute_parallel_for_codegen.cpp

// 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=CHECK2
// 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=CHECK4

// 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 --implicit-check-not="{{__kmpc|__tgt}}"
// 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 --implicit-check-not="{{__kmpc|__tgt}}"
// 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 --implicit-check-not="{{__kmpc|__tgt}}"
// 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 --implicit-check-not="{{__kmpc|__tgt}}"

// 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-target-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-target-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-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK10
// 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-target-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-target-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-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK12

// 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-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// 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-target-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-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// 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-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// 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-target-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-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"

#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: (ptr noundef [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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__CASTED2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    store i32 1000, ptr [[N]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK1-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
// CHECK1-NEXT:    store i32 128, ptr [[TH]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 8
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51(i64 [[TMP4]], ptr [[A]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2:[0-9]+]]
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[N]], align 4
// CHECK1-NEXT:    store i32 [[TMP9]], ptr [[N_CASTED3]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i64, ptr [[N_CASTED3]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58(i64 [[TMP10]], ptr [[A]], ptr [[TMP11]]) #[[ATTR2]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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(ptr @[[GLOB3:[0-9]+]])
// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK1-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined, i64 [[TMP5]], ptr [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       omp.precond.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[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]], ptr [[I4]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I4]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(ptr @[[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, ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP24]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       cancel.exit:
// CHECK1-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], 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 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined, i64 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], ptr [[TMP20]])
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP24]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       omp.precond.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, align 8
// CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[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]], ptr [[I4]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i64 0
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 [[TMP18]], ptr [[ARRAYIDX7]], 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, ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK1-NEXT:    store i32 [[ADD8]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], 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: (ptr noundef [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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__CASTED2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK2-NEXT:    [[N_CASTED7:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [3 x ptr], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [3 x ptr], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [3 x ptr], align 8
// CHECK2-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[KERNEL_ARGS18:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK2-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK2-NEXT:    store i32 1000, ptr [[N]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK2-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
// CHECK2-NEXT:    store i32 128, ptr [[TH]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
// CHECK2-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
// CHECK2-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 8
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    store i64 [[TMP4]], ptr [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    store i64 [[TMP4]], ptr [[TMP10]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store ptr null, ptr [[TMP11]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    store ptr [[A]], ptr [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    store ptr [[A]], ptr [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store ptr null, ptr [[TMP14]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    store i64 [[TMP6]], ptr [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    store i64 [[TMP6]], ptr [[TMP16]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK2-NEXT:    store i64 [[TMP8]], ptr [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK2-NEXT:    store i64 [[TMP8]], ptr [[TMP19]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK2-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, ptr [[N]], align 4
// CHECK2-NEXT:    store i32 [[TMP25]], ptr [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK2-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB]], 1
// CHECK2-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[DIV5]], 1
// CHECK2-NEXT:    store i32 [[SUB6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP27]], 1
// CHECK2-NEXT:    [[TMP28:%.*]] = zext i32 [[ADD]] to i64
// CHECK2-NEXT:    [[TMP29:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP23]], 0
// CHECK2-NEXT:    [[TMP30:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP24]], 0
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK2-NEXT:    store i32 3, ptr [[TMP31]], align 4
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK2-NEXT:    store i32 4, ptr [[TMP32]], align 4
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK2-NEXT:    store ptr [[TMP21]], ptr [[TMP33]], align 8
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK2-NEXT:    store ptr [[TMP22]], ptr [[TMP34]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK2-NEXT:    store ptr @.offload_sizes, ptr [[TMP35]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK2-NEXT:    store ptr @.offload_maptypes, ptr [[TMP36]], align 8
// CHECK2-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK2-NEXT:    store ptr null, ptr [[TMP37]], align 8
// CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK2-NEXT:    store ptr null, ptr [[TMP38]], align 8
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK2-NEXT:    store i64 [[TMP28]], ptr [[TMP39]], align 8
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK2-NEXT:    store i64 0, ptr [[TMP40]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK2-NEXT:    store [3 x i32] [[TMP29]], ptr [[TMP41]], align 4
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK2-NEXT:    store [3 x i32] [[TMP30]], ptr [[TMP42]], align 4
// CHECK2-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK2-NEXT:    store i32 0, ptr [[TMP43]], align 4
// CHECK2-NEXT:    [[TMP44:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 [[TMP23]], i32 [[TMP24]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.region_id, ptr [[KERNEL_ARGS]])
// CHECK2-NEXT:    [[TMP45:%.*]] = icmp ne i32 [[TMP44]], 0
// CHECK2-NEXT:    br i1 [[TMP45]], 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]], ptr [[A]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2:[0-9]+]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    [[TMP46:%.*]] = load i32, ptr [[N]], align 4
// CHECK2-NEXT:    store i32 [[TMP46]], ptr [[N_CASTED7]], align 4
// CHECK2-NEXT:    [[TMP47:%.*]] = load i64, ptr [[N_CASTED7]], align 8
// CHECK2-NEXT:    [[TMP48:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK2-NEXT:    store i64 [[TMP47]], ptr [[TMP49]], align 8
// CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK2-NEXT:    store i64 [[TMP47]], ptr [[TMP50]], align 8
// CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
// CHECK2-NEXT:    store ptr null, ptr [[TMP51]], align 8
// CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK2-NEXT:    store ptr [[A]], ptr [[TMP52]], align 8
// CHECK2-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK2-NEXT:    store ptr [[A]], ptr [[TMP53]], align 8
// CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
// CHECK2-NEXT:    store ptr null, ptr [[TMP54]], align 8
// CHECK2-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 2
// CHECK2-NEXT:    store ptr [[TMP48]], ptr [[TMP55]], align 8
// CHECK2-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 2
// CHECK2-NEXT:    store ptr [[TMP48]], ptr [[TMP56]], align 8
// CHECK2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 2
// CHECK2-NEXT:    store ptr null, ptr [[TMP57]], align 8
// CHECK2-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP60:%.*]] = load i32, ptr [[N]], align 4
// CHECK2-NEXT:    store i32 [[TMP60]], ptr [[DOTCAPTURE_EXPR_12]], align 4
// CHECK2-NEXT:    [[TMP61:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_12]], align 4
// CHECK2-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP61]], 0
// CHECK2-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
// CHECK2-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
// CHECK2-NEXT:    store i32 [[SUB16]], ptr [[DOTCAPTURE_EXPR_13]], align 4
// CHECK2-NEXT:    [[TMP62:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_13]], align 4
// CHECK2-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP62]], 1
// CHECK2-NEXT:    [[TMP63:%.*]] = zext i32 [[ADD17]] to i64
// CHECK2-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 0
// CHECK2-NEXT:    store i32 3, ptr [[TMP64]], align 4
// CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 1
// CHECK2-NEXT:    store i32 3, ptr [[TMP65]], align 4
// CHECK2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 2
// CHECK2-NEXT:    store ptr [[TMP58]], ptr [[TMP66]], align 8
// CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 3
// CHECK2-NEXT:    store ptr [[TMP59]], ptr [[TMP67]], align 8
// CHECK2-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 4
// CHECK2-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP68]], align 8
// CHECK2-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 5
// CHECK2-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP69]], align 8
// CHECK2-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 6
// CHECK2-NEXT:    store ptr null, ptr [[TMP70]], align 8
// CHECK2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 7
// CHECK2-NEXT:    store ptr null, ptr [[TMP71]], align 8
// CHECK2-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 8
// CHECK2-NEXT:    store i64 [[TMP63]], ptr [[TMP72]], align 8
// CHECK2-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 9
// CHECK2-NEXT:    store i64 0, ptr [[TMP73]], align 8
// CHECK2-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 10
// CHECK2-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP74]], align 4
// CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 11
// CHECK2-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP75]], align 4
// CHECK2-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 12
// CHECK2-NEXT:    store i32 0, ptr [[TMP76]], align 4
// CHECK2-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.region_id, ptr [[KERNEL_ARGS18]])
// CHECK2-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK2-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
// CHECK2:       omp_offload.failed19:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58(i64 [[TMP47]], ptr [[A]], ptr [[TMP48]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
// CHECK2:       omp_offload.cont20:
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP79:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK2-NEXT:    ret i32 [[TMP79]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK2-SAME: (i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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(ptr @[[GLOB3]])
// CHECK2-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK2-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK2-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined, i64 [[TMP5]], ptr [[TMP1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined
// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK2-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]])
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK2-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined
// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[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]], ptr [[I4]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I4]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(ptr @[[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, ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[ADD7]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP24]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       cancel.exit:
// CHECK2-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], 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 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK2-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined, i64 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined
// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK2-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], ptr [[TMP20]])
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK2-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP24]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined
// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, align 8
// CHECK2-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[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]], ptr [[I4]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i64 0
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 [[TMP18]], ptr [[ARRAYIDX7]], 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, ptr [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK2-NEXT:    store i32 [[ADD8]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK2:       omp.precond.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
// CHECK4-SAME: (ptr noundef [[G:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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 ptr], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], 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:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK4-NEXT:    [[N_CASTED7:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [3 x ptr], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [3 x ptr], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [3 x ptr], 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:    [[KERNEL_ARGS18:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK4-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
// CHECK4-NEXT:    store i32 1000, ptr [[N]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
// CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
// CHECK4-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
// CHECK4-NEXT:    store i32 128, ptr [[TH]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
// CHECK4-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
// CHECK4-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    store i32 [[TMP4]], ptr [[TMP9]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    store i32 [[TMP4]], ptr [[TMP10]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store ptr null, ptr [[TMP11]], align 4
// CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    store ptr [[A]], ptr [[TMP12]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    store ptr [[A]], ptr [[TMP13]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store ptr null, ptr [[TMP14]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    store i32 [[TMP6]], ptr [[TMP15]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    store i32 [[TMP6]], ptr [[TMP16]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store ptr null, ptr [[TMP17]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK4-NEXT:    store i32 [[TMP8]], ptr [[TMP18]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK4-NEXT:    store i32 [[TMP8]], ptr [[TMP19]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK4-NEXT:    store ptr null, ptr [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[TMP25:%.*]] = load i32, ptr [[N]], align 4
// CHECK4-NEXT:    store i32 [[TMP25]], ptr [[DOTCAPTURE_EXPR_3]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP26]], 0
// CHECK4-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB]], 1
// CHECK4-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[DIV5]], 1
// CHECK4-NEXT:    store i32 [[SUB6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
// CHECK4-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP27]], 1
// CHECK4-NEXT:    [[TMP28:%.*]] = zext i32 [[ADD]] to i64
// CHECK4-NEXT:    [[TMP29:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP23]], 0
// CHECK4-NEXT:    [[TMP30:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP24]], 0
// CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK4-NEXT:    store i32 3, ptr [[TMP31]], align 4
// CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK4-NEXT:    store i32 4, ptr [[TMP32]], align 4
// CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK4-NEXT:    store ptr [[TMP21]], ptr [[TMP33]], align 4
// CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK4-NEXT:    store ptr [[TMP22]], ptr [[TMP34]], align 4
// CHECK4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK4-NEXT:    store ptr @.offload_sizes, ptr [[TMP35]], align 4
// CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK4-NEXT:    store ptr @.offload_maptypes, ptr [[TMP36]], align 4
// CHECK4-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK4-NEXT:    store ptr null, ptr [[TMP37]], align 4
// CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK4-NEXT:    store ptr null, ptr [[TMP38]], align 4
// CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK4-NEXT:    store i64 [[TMP28]], ptr [[TMP39]], align 8
// CHECK4-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK4-NEXT:    store i64 0, ptr [[TMP40]], align 8
// CHECK4-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK4-NEXT:    store [3 x i32] [[TMP29]], ptr [[TMP41]], align 4
// CHECK4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK4-NEXT:    store [3 x i32] [[TMP30]], ptr [[TMP42]], align 4
// CHECK4-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK4-NEXT:    store i32 0, ptr [[TMP43]], align 4
// CHECK4-NEXT:    [[TMP44:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 [[TMP23]], i32 [[TMP24]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.region_id, ptr [[KERNEL_ARGS]])
// CHECK4-NEXT:    [[TMP45:%.*]] = icmp ne i32 [[TMP44]], 0
// CHECK4-NEXT:    br i1 [[TMP45]], 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]], ptr [[A]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2:[0-9]+]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    [[TMP46:%.*]] = load i32, ptr [[N]], align 4
// CHECK4-NEXT:    store i32 [[TMP46]], ptr [[N_CASTED7]], align 4
// CHECK4-NEXT:    [[TMP47:%.*]] = load i32, ptr [[N_CASTED7]], align 4
// CHECK4-NEXT:    [[TMP48:%.*]] = load ptr, ptr [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK4-NEXT:    store i32 [[TMP47]], ptr [[TMP49]], align 4
// CHECK4-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK4-NEXT:    store i32 [[TMP47]], ptr [[TMP50]], align 4
// CHECK4-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
// CHECK4-NEXT:    store ptr null, ptr [[TMP51]], align 4
// CHECK4-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
// CHECK4-NEXT:    store ptr [[A]], ptr [[TMP52]], align 4
// CHECK4-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
// CHECK4-NEXT:    store ptr [[A]], ptr [[TMP53]], align 4
// CHECK4-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
// CHECK4-NEXT:    store ptr null, ptr [[TMP54]], align 4
// CHECK4-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 2
// CHECK4-NEXT:    store ptr [[TMP48]], ptr [[TMP55]], align 4
// CHECK4-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 2
// CHECK4-NEXT:    store ptr [[TMP48]], ptr [[TMP56]], align 4
// CHECK4-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 2
// CHECK4-NEXT:    store ptr null, ptr [[TMP57]], align 4
// CHECK4-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP60:%.*]] = load i32, ptr [[N]], align 4
// CHECK4-NEXT:    store i32 [[TMP60]], ptr [[DOTCAPTURE_EXPR_12]], align 4
// CHECK4-NEXT:    [[TMP61:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_12]], align 4
// CHECK4-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP61]], 0
// CHECK4-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
// CHECK4-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
// CHECK4-NEXT:    store i32 [[SUB16]], ptr [[DOTCAPTURE_EXPR_13]], align 4
// CHECK4-NEXT:    [[TMP62:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_13]], align 4
// CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP62]], 1
// CHECK4-NEXT:    [[TMP63:%.*]] = zext i32 [[ADD17]] to i64
// CHECK4-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 0
// CHECK4-NEXT:    store i32 3, ptr [[TMP64]], align 4
// CHECK4-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 1
// CHECK4-NEXT:    store i32 3, ptr [[TMP65]], align 4
// CHECK4-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 2
// CHECK4-NEXT:    store ptr [[TMP58]], ptr [[TMP66]], align 4
// CHECK4-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 3
// CHECK4-NEXT:    store ptr [[TMP59]], ptr [[TMP67]], align 4
// CHECK4-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 4
// CHECK4-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP68]], align 4
// CHECK4-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 5
// CHECK4-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP69]], align 4
// CHECK4-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 6
// CHECK4-NEXT:    store ptr null, ptr [[TMP70]], align 4
// CHECK4-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 7
// CHECK4-NEXT:    store ptr null, ptr [[TMP71]], align 4
// CHECK4-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 8
// CHECK4-NEXT:    store i64 [[TMP63]], ptr [[TMP72]], align 8
// CHECK4-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 9
// CHECK4-NEXT:    store i64 0, ptr [[TMP73]], align 8
// CHECK4-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 10
// CHECK4-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP74]], align 4
// CHECK4-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 11
// CHECK4-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP75]], align 4
// CHECK4-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 12
// CHECK4-NEXT:    store i32 0, ptr [[TMP76]], align 4
// CHECK4-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.region_id, ptr [[KERNEL_ARGS18]])
// CHECK4-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK4-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
// CHECK4:       omp_offload.failed19:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58(i32 [[TMP47]], ptr [[A]], ptr [[TMP48]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
// CHECK4:       omp_offload.cont20:
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP79:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK4-NEXT:    ret i32 [[TMP79]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK4-SAME: (i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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(ptr @[[GLOB3]])
// CHECK4-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK4-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined, i32 [[TMP5]], ptr [[TMP1]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined
// CHECK4-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP16]], ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined, i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], ptr [[TMP0]])
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK4-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP21]])
// CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK4:       omp.precond.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined
// CHECK4-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, ptr [[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, ptr [[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]], ptr [[I3]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 [[TMP17]]
// CHECK4-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(ptr @[[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, ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK4-NEXT:    store i32 [[ADD6]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP24]])
// CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK4:       cancel.exit:
// CHECK4-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], 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 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 4
// CHECK4-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined, i32 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined
// CHECK4-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP16]], ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[G_ADDR]], align 4
// CHECK4-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined, i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], ptr [[TMP0]], ptr [[TMP18]])
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK4-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
// CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK4:       omp.precond.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined
// CHECK4-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, align 4
// CHECK4-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, ptr [[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, ptr [[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]], ptr [[I3]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 4
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I3]], align 4
// CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 [[TMP19]]
// CHECK4-NEXT:    store i32 [[TMP18]], ptr [[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, ptr [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK4-NEXT:    store i32 [[ADD7]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
// CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK4:       omp.precond.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK10-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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(ptr @[[GLOB3:[0-9]+]])
// CHECK10-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK10-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK10-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined, i64 [[TMP5]], ptr [[TMP1]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined
// CHECK10-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK10-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]])
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
// CHECK10-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
// CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK10:       omp.precond.end:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined
// CHECK10-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK10-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK10-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK10-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK10-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, ptr [[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]], ptr [[I4]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I4]], align 4
// CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK10-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
// CHECK10-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(ptr @[[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, ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK10-NEXT:    store i32 [[ADD7]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP24]])
// CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK10:       cancel.exit:
// CHECK10-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], 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: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK10-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined, i64 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined
// CHECK10-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = load i64, ptr [[N_CASTED]], align 8
// CHECK10-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK10-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], ptr [[TMP20]])
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK10:       omp.inner.for.inc:
// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
// CHECK10-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP24]])
// CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK10:       omp.precond.end:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined
// CHECK10-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, align 8
// CHECK10-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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:    [[I4:%.*]] = alloca i32, align 4
// CHECK10-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK10-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
// CHECK10-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
// CHECK10-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK10-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK10-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
// CHECK10-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK10:       cond.true:
// CHECK10-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK10-NEXT:    br label [[COND_END:%.*]]
// CHECK10:       cond.false:
// CHECK10-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK10-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK10:       omp.inner.for.cond:
// CHECK10-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK10-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK10-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK10:       omp.inner.for.body:
// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, ptr [[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]], ptr [[I4]], align 4
// CHECK10-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i64 0
// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4
// CHECK10-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK10-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
// CHECK10-NEXT:    store i32 [[TMP18]], ptr [[ARRAYIDX7]], 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, ptr [[DOTOMP_IV]], align 4
// CHECK10-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK10-NEXT:    store i32 [[ADD8]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
// CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
// CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK10:       omp.precond.end:
// CHECK10-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51
// CHECK12-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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(ptr @[[GLOB3:[0-9]+]])
// CHECK12-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK12-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK12-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined, i32 [[TMP5]], ptr [[TMP1]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined
// CHECK12-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK12-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    br label [[COND_END:%.*]]
// CHECK12:       cond.false:
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP16]], ptr [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP17:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK12-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined, i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], ptr [[TMP0]])
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK12:       omp.inner.for.inc:
// CHECK12-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
// CHECK12-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP21]])
// CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK12:       omp.precond.end:
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l51.omp_outlined.omp_outlined
// CHECK12-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK12-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK12-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    br label [[COND_END:%.*]]
// CHECK12:       cond.false:
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, ptr [[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, ptr [[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]], ptr [[I3]], align 4
// CHECK12-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 [[TMP17]]
// CHECK12-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
// CHECK12-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
// CHECK12-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(ptr @[[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, ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK12-NEXT:    store i32 [[ADD6]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP24]])
// CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK12:       cancel.exit:
// CHECK12-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], 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: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 4
// CHECK12-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined, i32 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined
// CHECK12-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK12-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    br label [[COND_END:%.*]]
// CHECK12:       cond.false:
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK12-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP16]], ptr [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP17:%.*]] = load i32, ptr [[N_CASTED]], align 4
// CHECK12-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[G_ADDR]], align 4
// CHECK12-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined, i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], ptr [[TMP0]], ptr [[TMP18]])
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK12:       omp.inner.for.inc:
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
// CHECK12-NEXT:    store i32 [[ADD]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
// CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK12:       omp.precond.end:
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l58.omp_outlined.omp_outlined
// CHECK12-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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 ptr, align 4
// CHECK12-NEXT:    [[G_ADDR:%.*]] = alloca ptr, 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 ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, ptr [[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]], ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 0, ptr [[I]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, ptr [[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, ptr [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK12-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK12-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK12-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP10:%.*]] = load i32, ptr [[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, ptr [[DOTCAPTURE_EXPR_1]], align 4
// CHECK12-NEXT:    br label [[COND_END:%.*]]
// CHECK12:       cond.false:
// CHECK12-NEXT:    [[TMP12:%.*]] = load i32, ptr [[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]], ptr [[DOTOMP_UB]], align 4
// CHECK12-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK12-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK12:       omp.inner.for.cond:
// CHECK12-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[TMP15:%.*]] = load i32, ptr [[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, ptr [[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]], ptr [[I3]], align 4
// CHECK12-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 4
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
// CHECK12-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
// CHECK12-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I3]], align 4
// CHECK12-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 [[TMP19]]
// CHECK12-NEXT:    store i32 [[TMP18]], ptr [[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, ptr [[DOTOMP_IV]], align 4
// CHECK12-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK12-NEXT:    store i32 [[ADD7]], ptr [[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 ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
// CHECK12-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
// CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK12:       omp.precond.end:
// CHECK12-NEXT:    ret void
//