// 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 _
// Test host codegen.
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK3
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK3

// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -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 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -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 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -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 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -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.
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11

// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"

// Test host codegen.
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK3
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK3

// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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.
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11

// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
// RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"

// expected-no-diagnostics
#ifndef HEADER
#define HEADER




// We have 8 target regions, but only 7 that actually will generate offloading
// code, only 6 will have mapped arguments, and only 4 have all-constant map
// sizes.



// Check target registration is registered as a Ctor.


template<typename tx, typename ty>
struct TT{
  tx X;
  ty Y;
};

int global;

int foo(int n) {
  int a = 0;
  short aa = 0;
  float b[10];
  float bn[n];
  double c[5][10];
  double cn[5][n];
  TT<long long, char> d;

  #pragma omp target teams distribute num_teams(a) thread_limit(a) firstprivate(aa) nowait
  for (int i = 0; i < 10; ++i) {
  }

  #pragma omp target teams distribute if(target: 0)
  for (int i = 0; i < 10; ++i) {
    a += 1;
  }


  #pragma omp target teams distribute if(target: 1)
  for (int i = 0; i < 10; ++i) {
    aa += 1;
  }



  #pragma omp target teams distribute if(target: n>10)
  for (int i = 0; i < 10; ++i) {
    a += 1;
    aa += 1;
  }

  // We capture 3 VLA sizes in this target region





  // The names below are not necessarily consistent with the names used for the
  // addresses above as some are repeated.











  #pragma omp target teams distribute if(target: n>20) dist_schedule(static, n)
  for (int i = 0; i < 10; ++i) {
    a += 1;
    b[2] += 1.0;
    bn[3] += 1.0;
    c[1][2] += 1.0;
    cn[1][3] += 1.0;
    d.X += 1;
    d.Y += 1;
  }

  return a;
}

// Check that the offloading functions are emitted and that the arguments are
// correct and loaded correctly for the target regions in foo().




// Create stack storage and store argument in there.

// Create stack storage and store argument in there.

// Create stack storage and store argument in there.

// Create local storage for each capture.



// To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.

template<typename tx>
tx ftemplate(int n) {
  tx a = 0;
  short aa = 0;
  tx b[10];

  #pragma omp target teams distribute if(target: n>40)
  for (int i = 0; i < 10; ++i) {
    a += 1;
    aa += 1;
    b[2] += 1;
  }

  return a;
}

static
int fstatic(int n) {
  int a = 0;
  short aa = 0;
  char aaa = 0;
  int b[10];

  #pragma omp target teams distribute if(target: n>50)
  for (int i = a; i < n; ++i) {
    a += 1;
    aa += 1;
    aaa += 1;
    b[2] += 1;
  }

  return a;
}

struct S1 {
  double a;

  int r1(int n){
    int b = n+1;
    short int c[2][n];

    #pragma omp target teams distribute if(target: n>60)
    for (int i = 0; i < 10; ++i) {
      this->a = (double)b + 1.5;
      c[1][1] = ++a;
    }

    return c[1][1] + (int)b;
  }
};

int bar(int n){
  int a = 0;

  a += foo(n);

  S1 S;
  a += S.r1(n);

  a += fstatic(n);

  a += ftemplate<int>(n);

  return a;
}



// We capture 2 VLA sizes in this target region


// The names below are not necessarily consistent with the names used for the
// addresses above as some are repeated.





















// Check that the offloading functions are emitted and that the arguments are
// correct and loaded correctly for the target regions of the callees of bar().

// Create local storage for each capture.
// Store captures in the context.


// To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.


// Create local storage for each capture.
// Store captures in the context.




// To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.

// Create local storage for each capture.
// Store captures in the context.



// To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.

#endif
// CHECK1-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
// CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
// CHECK1-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    [[_TMP19:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[A_CASTED24:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED26:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS30:%.*]] = alloca [10 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS31:%.*]] = alloca [10 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS32:%.*]] = alloca [10 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 8
// CHECK1-NEXT:    [[_TMP33:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
// CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
// CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
// CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
// CHECK1-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK1-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
// CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
// CHECK1-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
// CHECK1-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK1-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
// CHECK1-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK1-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
// CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
// CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK1-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
// CHECK1-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i64 120, i64 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK1-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
// CHECK1-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
// CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
// CHECK1-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
// CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
// CHECK1-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
// CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP53]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
// CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
// CHECK1-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP38]])
// CHECK1-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
// CHECK1-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
// CHECK1-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
// CHECK1-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
// CHECK1-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
// CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
// CHECK1-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
// CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
// CHECK1-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
// CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP65]], align 8
// CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP68]], align 4
// CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK1-NEXT:    store i32 1, i32* [[TMP69]], align 4
// CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP66]], i8*** [[TMP70]], align 8
// CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP67]], i8*** [[TMP71]], align 8
// CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP72]], align 8
// CHECK1-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP73]], align 8
// CHECK1-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP74]], align 8
// CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP75]], align 8
// CHECK1-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK1-NEXT:    store i64 10, i64* [[TMP76]], align 8
// CHECK1-NEXT:    [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK1-NEXT:    [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
// CHECK1-NEXT:    br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i64 [[TMP60]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[TMP79:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
// CHECK1-NEXT:    store i32 [[TMP79]], i32* [[CONV13]], align 4
// CHECK1-NEXT:    [[TMP80:%.*]] = load i64, i64* [[A_CASTED12]], align 8
// CHECK1-NEXT:    [[TMP81:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
// CHECK1-NEXT:    store i16 [[TMP81]], i16* [[CONV15]], align 2
// CHECK1-NEXT:    [[TMP82:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
// CHECK1-NEXT:    [[TMP83:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP83]], 10
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i64*
// CHECK1-NEXT:    store i64 [[TMP80]], i64* [[TMP85]], align 8
// CHECK1-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i64*
// CHECK1-NEXT:    store i64 [[TMP80]], i64* [[TMP87]], align 8
// CHECK1-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP88]], align 8
// CHECK1-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP90:%.*]] = bitcast i8** [[TMP89]] to i64*
// CHECK1-NEXT:    store i64 [[TMP82]], i64* [[TMP90]], align 8
// CHECK1-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to i64*
// CHECK1-NEXT:    store i64 [[TMP82]], i64* [[TMP92]], align 8
// CHECK1-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP93]], align 8
// CHECK1-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS20:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP96]], align 4
// CHECK1-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 1
// CHECK1-NEXT:    store i32 2, i32* [[TMP97]], align 4
// CHECK1-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP94]], i8*** [[TMP98]], align 8
// CHECK1-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP95]], i8*** [[TMP99]], align 8
// CHECK1-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64** [[TMP100]], align 8
// CHECK1-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i64** [[TMP101]], align 8
// CHECK1-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP102]], align 8
// CHECK1-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP103]], align 8
// CHECK1-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]], i32 0, i32 8
// CHECK1-NEXT:    store i64 10, i64* [[TMP104]], align 8
// CHECK1-NEXT:    [[TMP105:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS20]])
// CHECK1-NEXT:    [[TMP106:%.*]] = icmp ne i32 [[TMP105]], 0
// CHECK1-NEXT:    br i1 [[TMP106]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
// CHECK1:       omp_offload.failed21:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP80]], i64 [[TMP82]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT22]]
// CHECK1:       omp_offload.cont22:
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP80]], i64 [[TMP82]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP107:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP107]], i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK1-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV25:%.*]] = bitcast i64* [[A_CASTED24]] to i32*
// CHECK1-NEXT:    store i32 [[TMP108]], i32* [[CONV25]], align 4
// CHECK1-NEXT:    [[TMP109:%.*]] = load i64, i64* [[A_CASTED24]], align 8
// CHECK1-NEXT:    [[TMP110:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
// CHECK1-NEXT:    [[CONV27:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED26]] to i32*
// CHECK1-NEXT:    store i32 [[TMP110]], i32* [[CONV27]], align 4
// CHECK1-NEXT:    [[TMP111:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED26]], align 8
// CHECK1-NEXT:    [[TMP112:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CMP28:%.*]] = icmp sgt i32 [[TMP112]], 20
// CHECK1-NEXT:    br i1 [[CMP28]], label [[OMP_IF_THEN29:%.*]], label [[OMP_IF_ELSE37:%.*]]
// CHECK1:       omp_if.then29:
// CHECK1-NEXT:    [[TMP113:%.*]] = mul nuw i64 [[TMP2]], 4
// CHECK1-NEXT:    [[TMP114:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK1-NEXT:    [[TMP115:%.*]] = mul nuw i64 [[TMP114]], 8
// CHECK1-NEXT:    [[TMP116:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP116]], i8* align 8 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i64 80, i1 false)
// CHECK1-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to i64*
// CHECK1-NEXT:    store i64 [[TMP109]], i64* [[TMP118]], align 8
// CHECK1-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i64*
// CHECK1-NEXT:    store i64 [[TMP109]], i64* [[TMP120]], align 8
// CHECK1-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP121]], align 8
// CHECK1-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [10 x float]**
// CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP123]], align 8
// CHECK1-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to [10 x float]**
// CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP125]], align 8
// CHECK1-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP126]], align 8
// CHECK1-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP128]], align 8
// CHECK1-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP130]], align 8
// CHECK1-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP131]], align 8
// CHECK1-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to float**
// CHECK1-NEXT:    store float* [[VLA]], float** [[TMP133]], align 8
// CHECK1-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to float**
// CHECK1-NEXT:    store float* [[VLA]], float** [[TMP135]], align 8
// CHECK1-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK1-NEXT:    store i64 [[TMP113]], i64* [[TMP136]], align 8
// CHECK1-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 3
// CHECK1-NEXT:    store i8* null, i8** [[TMP137]], align 8
// CHECK1-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to [5 x [10 x double]]**
// CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP139]], align 8
// CHECK1-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to [5 x [10 x double]]**
// CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP141]], align 8
// CHECK1-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP142]], align 8
// CHECK1-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to i64*
// CHECK1-NEXT:    store i64 5, i64* [[TMP144]], align 8
// CHECK1-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP146:%.*]] = bitcast i8** [[TMP145]] to i64*
// CHECK1-NEXT:    store i64 5, i64* [[TMP146]], align 8
// CHECK1-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 5
// CHECK1-NEXT:    store i8* null, i8** [[TMP147]], align 8
// CHECK1-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP149]], align 8
// CHECK1-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP151]], align 8
// CHECK1-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 6
// CHECK1-NEXT:    store i8* null, i8** [[TMP152]], align 8
// CHECK1-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP154:%.*]] = bitcast i8** [[TMP153]] to double**
// CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP154]], align 8
// CHECK1-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP156:%.*]] = bitcast i8** [[TMP155]] to double**
// CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP156]], align 8
// CHECK1-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK1-NEXT:    store i64 [[TMP115]], i64* [[TMP157]], align 8
// CHECK1-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 7
// CHECK1-NEXT:    store i8* null, i8** [[TMP158]], align 8
// CHECK1-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP160:%.*]] = bitcast i8** [[TMP159]] to %struct.TT**
// CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP160]], align 8
// CHECK1-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to %struct.TT**
// CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP162]], align 8
// CHECK1-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 8
// CHECK1-NEXT:    store i8* null, i8** [[TMP163]], align 8
// CHECK1-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 9
// CHECK1-NEXT:    [[TMP165:%.*]] = bitcast i8** [[TMP164]] to i64*
// CHECK1-NEXT:    store i64 [[TMP111]], i64* [[TMP165]], align 8
// CHECK1-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 9
// CHECK1-NEXT:    [[TMP167:%.*]] = bitcast i8** [[TMP166]] to i64*
// CHECK1-NEXT:    store i64 [[TMP111]], i64* [[TMP167]], align 8
// CHECK1-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS32]], i64 0, i64 9
// CHECK1-NEXT:    store i8* null, i8** [[TMP168]], align 8
// CHECK1-NEXT:    [[TMP169:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS30]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS31]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP171:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS34:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP172]], align 4
// CHECK1-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 1
// CHECK1-NEXT:    store i32 10, i32* [[TMP173]], align 4
// CHECK1-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP169]], i8*** [[TMP174]], align 8
// CHECK1-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP170]], i8*** [[TMP175]], align 8
// CHECK1-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 4
// CHECK1-NEXT:    store i64* [[TMP171]], i64** [[TMP176]], align 8
// CHECK1-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i64** [[TMP177]], align 8
// CHECK1-NEXT:    [[TMP178:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP178]], align 8
// CHECK1-NEXT:    [[TMP179:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP179]], align 8
// CHECK1-NEXT:    [[TMP180:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]], i32 0, i32 8
// CHECK1-NEXT:    store i64 10, i64* [[TMP180]], align 8
// CHECK1-NEXT:    [[TMP181:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS34]])
// CHECK1-NEXT:    [[TMP182:%.*]] = icmp ne i32 [[TMP181]], 0
// CHECK1-NEXT:    br i1 [[TMP182]], label [[OMP_OFFLOAD_FAILED35:%.*]], label [[OMP_OFFLOAD_CONT36:%.*]]
// CHECK1:       omp_offload.failed35:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP109]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP111]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT36]]
// CHECK1:       omp_offload.cont36:
// CHECK1-NEXT:    br label [[OMP_IF_END38:%.*]]
// CHECK1:       omp_if.else37:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP109]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP111]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_IF_END38]]
// CHECK1:       omp_if.end38:
// CHECK1-NEXT:    [[TMP183:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[TMP184:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP184]])
// CHECK1-NEXT:    ret i32 [[TMP183]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
// CHECK1-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
// CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], 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:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
// CHECK1-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
// CHECK1-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
// CHECK1-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
// CHECK1-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
// CHECK1-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
// CHECK1-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
// CHECK1-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
// CHECK1-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
// CHECK1-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
// CHECK1-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
// CHECK1-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
// CHECK1-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK1-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
// CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
// CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
// CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
// CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
// CHECK1-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[KERNEL_ARGS_I:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
// CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
// CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
// CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
// CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
// CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
// CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
// CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
// CHECK1-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
// CHECK1-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP27]], align 4, !noalias !21
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 1
// CHECK1-NEXT:    store i32 3, i32* [[TMP28]], align 4, !noalias !21
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP20]], i8*** [[TMP29]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP21]], i8*** [[TMP30]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 4
// CHECK1-NEXT:    store i64* [[TMP22]], i64** [[TMP31]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP32]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP33]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP34]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 8
// CHECK1-NEXT:    store i64 10, i64* [[TMP35]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel_nowait(%struct.ident_t* @[[GLOB2]], i64 -1, i32 [[TMP25]], i32 [[TMP26]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i8* null, i32 0, i8* null)
// CHECK1-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK1-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK1:       omp_offload.failed.i:
// CHECK1-NEXT:    [[TMP38:%.*]] = load i16, i16* [[TMP16]], align 2
// CHECK1-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
// CHECK1-NEXT:    store i16 [[TMP38]], i16* [[CONV_I]], align 2, !noalias !21
// CHECK1-NEXT:    [[TMP39:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
// CHECK1-NEXT:    store i32 [[TMP40]], i32* [[CONV4_I]], align 4, !noalias !21
// CHECK1-NEXT:    [[TMP41:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP42:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK1-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
// CHECK1-NEXT:    store i32 [[TMP42]], i32* [[CONV6_I]], align 4, !noalias !21
// CHECK1-NEXT:    [[TMP43:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !21
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i64 [[TMP39]], i64 [[TMP41]], i64 [[TMP43]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
// CHECK1:       .omp_outlined..1.exit:
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
// CHECK1-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[CONV]], 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:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
// CHECK1-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK1-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
// 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:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
// CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
// CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
// CHECK1-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
// CHECK1-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
// CHECK1-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
// 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:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
// CHECK1-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
// CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK1:       omp.dispatch.cond:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK1:       omp.dispatch.body:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22:![0-9]+]]
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
// CHECK1-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
// CHECK1-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
// CHECK1-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK1-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
// CHECK1-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
// CHECK1-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
// CHECK1-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK1-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
// CHECK1-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
// CHECK1-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
// CHECK1-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
// CHECK1-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
// CHECK1-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
// CHECK1-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group [[ACC_GRP22]]
// 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:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
// CHECK1-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK1:       omp.dispatch.inc:
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
// CHECK1-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
// CHECK1-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK1:       omp.dispatch.end:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z3bari
// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
// CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
// CHECK1-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
// CHECK1-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
// CHECK1-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
// CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
// CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
// CHECK1-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
// CHECK1-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
// CHECK1-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i64 40, i1 false)
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
// CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
// CHECK1-NEXT:    store double* [[A]], double** [[TMP14]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK1-NEXT:    store i64 2, i64* [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
// CHECK1-NEXT:    store i64 2, i64* [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK1-NEXT:    store i8* null, i8** [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
// CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
// CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP36]], align 8
// CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP40]], align 4
// CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK1-NEXT:    store i32 5, i32* [[TMP41]], align 4
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP37]], i8*** [[TMP42]], align 8
// CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP38]], i8*** [[TMP43]], align 8
// CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK1-NEXT:    store i64* [[TMP39]], i64** [[TMP44]], align 8
// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i64** [[TMP45]], align 8
// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP46]], align 8
// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP47]], align 8
// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK1-NEXT:    store i64 10, i64* [[TMP48]], align 8
// CHECK1-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK1-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
// CHECK1-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP51:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP51]]
// CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK1-NEXT:    [[TMP52:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP52]] to i32
// CHECK1-NEXT:    [[TMP53:%.*]] = load i32, i32* [[B]], align 4
// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP53]]
// CHECK1-NEXT:    [[TMP54:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP54]])
// CHECK1-NEXT:    ret i32 [[ADD4]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[AAA:%.*]] = alloca i8, align 1
// CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK1-NEXT:    store i8 0, i8* [[AAA]], align 1
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP4]], i16* [[CONV2]], align 2
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK1-NEXT:    store i8 [[TMP6]], i8* [[CONV3]], align 1
// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP17]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
// CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
// CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK1-NEXT:    store i8* null, i8** [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
// CHECK1-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
// CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
// CHECK1-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
// CHECK1-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add i32 [[TMP40]], 1
// CHECK1-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD8]] to i64
// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP42]], align 4
// CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK1-NEXT:    store i32 5, i32* [[TMP43]], align 4
// CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP34]], i8*** [[TMP44]], align 8
// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP35]], i8*** [[TMP45]], align 8
// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP46]], align 8
// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP47]], align 8
// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP48]], align 8
// CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP49]], align 8
// CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK1-NEXT:    store i64 [[TMP41]], i64* [[TMP50]], align 8
// CHECK1-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK1-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
// CHECK1-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP53:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    ret i32 [[TMP53]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
// CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP14]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK1-NEXT:    store i32 1, i32* [[TMP22]], align 4
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK1-NEXT:    store i32 3, i32* [[TMP23]], align 4
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK1-NEXT:    store i8** [[TMP20]], i8*** [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK1-NEXT:    store i8** [[TMP21]], i8*** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK1-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64** [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK1-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i64** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK1-NEXT:    store i8** null, i8*** [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK1-NEXT:    store i8** null, i8*** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK1-NEXT:    store i64 10, i64* [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK1-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK1-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    ret i32 [[TMP33]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
// CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
// CHECK1-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    store double [[ADD5]], double* [[A]], align 8
// CHECK1-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
// CHECK1-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK1-NEXT:    store double [[INC]], double* [[A6]], align 8
// CHECK1-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
// CHECK1-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
// CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK1-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
// CHECK1-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
// CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK1-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
// CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = 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:    [[I8:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
// CHECK1-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
// CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
// CHECK1-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
// CHECK1-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK1:       omp.precond.then:
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK1-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
// CHECK1-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
// CHECK1-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK1-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
// CHECK1-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
// CHECK1-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
// CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
// CHECK1-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
// CHECK1-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
// CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
// CHECK1-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
// CHECK1-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
// CHECK1-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK1-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
// CHECK1-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK1:       omp.precond.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
// CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
// CHECK1-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
// CHECK1-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
// CHECK1-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
// CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
// CHECK3-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
// CHECK3-NEXT:    [[_TMP15:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_CASTED20:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED21:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [10 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS25:%.*]] = alloca [10 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [10 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 4
// CHECK3-NEXT:    [[_TMP27:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
// CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK3-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK3-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK3-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
// CHECK3-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP22]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
// CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
// CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK3-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
// CHECK3-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 72, i32 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
// CHECK3-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
// CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
// CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
// CHECK3-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
// CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
// CHECK3-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
// CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
// CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
// CHECK3-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP36]])
// CHECK3-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
// CHECK3-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
// CHECK3-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
// CHECK3-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
// CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
// CHECK3-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
// CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK3-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
// CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP63]], align 4
// CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK3-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP66]], align 4
// CHECK3-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK3-NEXT:    store i32 1, i32* [[TMP67]], align 4
// CHECK3-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP64]], i8*** [[TMP68]], align 4
// CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP65]], i8*** [[TMP69]], align 4
// CHECK3-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP70]], align 4
// CHECK3-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP71]], align 4
// CHECK3-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP72]], align 4
// CHECK3-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP73]], align 4
// CHECK3-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK3-NEXT:    store i64 10, i64* [[TMP74]], align 8
// CHECK3-NEXT:    [[TMP75:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK3-NEXT:    [[TMP76:%.*]] = icmp ne i32 [[TMP75]], 0
// CHECK3-NEXT:    br i1 [[TMP76]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i32 [[TMP58]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    [[TMP77:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP77]], i32* [[A_CASTED9]], align 4
// CHECK3-NEXT:    [[TMP78:%.*]] = load i32, i32* [[A_CASTED9]], align 4
// CHECK3-NEXT:    [[TMP79:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
// CHECK3-NEXT:    store i16 [[TMP79]], i16* [[CONV11]], align 2
// CHECK3-NEXT:    [[TMP80:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
// CHECK3-NEXT:    [[TMP81:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP81]], 10
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
// CHECK3-NEXT:    store i32 [[TMP78]], i32* [[TMP83]], align 4
// CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
// CHECK3-NEXT:    store i32 [[TMP78]], i32* [[TMP85]], align 4
// CHECK3-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP86]], align 4
// CHECK3-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
// CHECK3-NEXT:    store i32 [[TMP80]], i32* [[TMP88]], align 4
// CHECK3-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP90:%.*]] = bitcast i8** [[TMP89]] to i32*
// CHECK3-NEXT:    store i32 [[TMP80]], i32* [[TMP90]], align 4
// CHECK3-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP91]], align 4
// CHECK3-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS16:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK3-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP94]], align 4
// CHECK3-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 1
// CHECK3-NEXT:    store i32 2, i32* [[TMP95]], align 4
// CHECK3-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP92]], i8*** [[TMP96]], align 4
// CHECK3-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP93]], i8*** [[TMP97]], align 4
// CHECK3-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64** [[TMP98]], align 4
// CHECK3-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i64** [[TMP99]], align 4
// CHECK3-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP100]], align 4
// CHECK3-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP101]], align 4
// CHECK3-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 8
// CHECK3-NEXT:    store i64 10, i64* [[TMP102]], align 8
// CHECK3-NEXT:    [[TMP103:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]])
// CHECK3-NEXT:    [[TMP104:%.*]] = icmp ne i32 [[TMP103]], 0
// CHECK3-NEXT:    br i1 [[TMP104]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
// CHECK3:       omp_offload.failed17:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP78]], i32 [[TMP80]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
// CHECK3:       omp_offload.cont18:
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP78]], i32 [[TMP80]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP105:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP105]], i32* [[DOTCAPTURE_EXPR_19]], align 4
// CHECK3-NEXT:    [[TMP106:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP106]], i32* [[A_CASTED20]], align 4
// CHECK3-NEXT:    [[TMP107:%.*]] = load i32, i32* [[A_CASTED20]], align 4
// CHECK3-NEXT:    [[TMP108:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
// CHECK3-NEXT:    store i32 [[TMP108]], i32* [[DOTCAPTURE_EXPR__CASTED21]], align 4
// CHECK3-NEXT:    [[TMP109:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED21]], align 4
// CHECK3-NEXT:    [[TMP110:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CMP22:%.*]] = icmp sgt i32 [[TMP110]], 20
// CHECK3-NEXT:    br i1 [[CMP22]], label [[OMP_IF_THEN23:%.*]], label [[OMP_IF_ELSE31:%.*]]
// CHECK3:       omp_if.then23:
// CHECK3-NEXT:    [[TMP111:%.*]] = mul nuw i32 [[TMP1]], 4
// CHECK3-NEXT:    [[TMP112:%.*]] = sext i32 [[TMP111]] to i64
// CHECK3-NEXT:    [[TMP113:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK3-NEXT:    [[TMP114:%.*]] = mul nuw i32 [[TMP113]], 8
// CHECK3-NEXT:    [[TMP115:%.*]] = sext i32 [[TMP114]] to i64
// CHECK3-NEXT:    [[TMP116:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP116]], i8* align 4 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i32 80, i1 false)
// CHECK3-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to i32*
// CHECK3-NEXT:    store i32 [[TMP107]], i32* [[TMP118]], align 4
// CHECK3-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32*
// CHECK3-NEXT:    store i32 [[TMP107]], i32* [[TMP120]], align 4
// CHECK3-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP121]], align 4
// CHECK3-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [10 x float]**
// CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP123]], align 4
// CHECK3-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to [10 x float]**
// CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP125]], align 4
// CHECK3-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP126]], align 4
// CHECK3-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP128]], align 4
// CHECK3-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP130]], align 4
// CHECK3-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP131]], align 4
// CHECK3-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to float**
// CHECK3-NEXT:    store float* [[VLA]], float** [[TMP133]], align 4
// CHECK3-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to float**
// CHECK3-NEXT:    store float* [[VLA]], float** [[TMP135]], align 4
// CHECK3-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK3-NEXT:    store i64 [[TMP112]], i64* [[TMP136]], align 4
// CHECK3-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP137]], align 4
// CHECK3-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to [5 x [10 x double]]**
// CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP139]], align 4
// CHECK3-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to [5 x [10 x double]]**
// CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP141]], align 4
// CHECK3-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 4
// CHECK3-NEXT:    store i8* null, i8** [[TMP142]], align 4
// CHECK3-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to i32*
// CHECK3-NEXT:    store i32 5, i32* [[TMP144]], align 4
// CHECK3-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP146:%.*]] = bitcast i8** [[TMP145]] to i32*
// CHECK3-NEXT:    store i32 5, i32* [[TMP146]], align 4
// CHECK3-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 5
// CHECK3-NEXT:    store i8* null, i8** [[TMP147]], align 4
// CHECK3-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP149]], align 4
// CHECK3-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP151]], align 4
// CHECK3-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 6
// CHECK3-NEXT:    store i8* null, i8** [[TMP152]], align 4
// CHECK3-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP154:%.*]] = bitcast i8** [[TMP153]] to double**
// CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP154]], align 4
// CHECK3-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP156:%.*]] = bitcast i8** [[TMP155]] to double**
// CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP156]], align 4
// CHECK3-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK3-NEXT:    store i64 [[TMP115]], i64* [[TMP157]], align 4
// CHECK3-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 7
// CHECK3-NEXT:    store i8* null, i8** [[TMP158]], align 4
// CHECK3-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP160:%.*]] = bitcast i8** [[TMP159]] to %struct.TT**
// CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP160]], align 4
// CHECK3-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to %struct.TT**
// CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP162]], align 4
// CHECK3-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 8
// CHECK3-NEXT:    store i8* null, i8** [[TMP163]], align 4
// CHECK3-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 9
// CHECK3-NEXT:    [[TMP165:%.*]] = bitcast i8** [[TMP164]] to i32*
// CHECK3-NEXT:    store i32 [[TMP109]], i32* [[TMP165]], align 4
// CHECK3-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 9
// CHECK3-NEXT:    [[TMP167:%.*]] = bitcast i8** [[TMP166]] to i32*
// CHECK3-NEXT:    store i32 [[TMP109]], i32* [[TMP167]], align 4
// CHECK3-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 9
// CHECK3-NEXT:    store i8* null, i8** [[TMP168]], align 4
// CHECK3-NEXT:    [[TMP169:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP171:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS28:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK3-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP172]], align 4
// CHECK3-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 1
// CHECK3-NEXT:    store i32 10, i32* [[TMP173]], align 4
// CHECK3-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP169]], i8*** [[TMP174]], align 4
// CHECK3-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP170]], i8*** [[TMP175]], align 4
// CHECK3-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 4
// CHECK3-NEXT:    store i64* [[TMP171]], i64** [[TMP176]], align 4
// CHECK3-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i64** [[TMP177]], align 4
// CHECK3-NEXT:    [[TMP178:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP178]], align 4
// CHECK3-NEXT:    [[TMP179:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP179]], align 4
// CHECK3-NEXT:    [[TMP180:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]], i32 0, i32 8
// CHECK3-NEXT:    store i64 10, i64* [[TMP180]], align 8
// CHECK3-NEXT:    [[TMP181:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS28]])
// CHECK3-NEXT:    [[TMP182:%.*]] = icmp ne i32 [[TMP181]], 0
// CHECK3-NEXT:    br i1 [[TMP182]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
// CHECK3:       omp_offload.failed29:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP107]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP109]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT30]]
// CHECK3:       omp_offload.cont30:
// CHECK3-NEXT:    br label [[OMP_IF_END32:%.*]]
// CHECK3:       omp_if.else31:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP107]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP109]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_IF_END32]]
// CHECK3:       omp_if.end32:
// CHECK3-NEXT:    [[TMP183:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    [[TMP184:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP184]])
// CHECK3-NEXT:    ret i32 [[TMP183]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
// CHECK3-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_task_privates_map.
// CHECK3-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
// CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
// CHECK3-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
// CHECK3-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
// CHECK3-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
// CHECK3-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
// CHECK3-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
// CHECK3-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
// CHECK3-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
// CHECK3-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
// CHECK3-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
// CHECK3-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
// CHECK3-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
// CHECK3-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK3-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
// CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
// CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
// CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
// CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
// CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
// CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
// CHECK3-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[KERNEL_ARGS_I:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
// CHECK3-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
// CHECK3-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
// CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
// CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
// CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
// CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
// CHECK3-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
// CHECK3-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP27]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 1
// CHECK3-NEXT:    store i32 3, i32* [[TMP28]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP20]], i8*** [[TMP29]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP21]], i8*** [[TMP30]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 4
// CHECK3-NEXT:    store i64* [[TMP22]], i64** [[TMP31]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP32]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP33]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP34]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i32 8
// CHECK3-NEXT:    store i64 10, i64* [[TMP35]], align 8, !noalias !22
// CHECK3-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel_nowait(%struct.ident_t* @[[GLOB2]], i64 -1, i32 [[TMP25]], i32 [[TMP26]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS_I]], i32 0, i8* null, i32 0, i8* null)
// CHECK3-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
// CHECK3-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK3:       omp_offload.failed.i:
// CHECK3-NEXT:    [[TMP38:%.*]] = load i16, i16* [[TMP16]], align 2
// CHECK3-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
// CHECK3-NEXT:    store i16 [[TMP38]], i16* [[CONV_I]], align 2, !noalias !22
// CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK3-NEXT:    store i32 [[TMP40]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP42:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK3-NEXT:    store i32 [[TMP42]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i32 [[TMP39]], i32 [[TMP41]], i32 [[TMP43]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
// CHECK3:       .omp_outlined..1.exit:
// CHECK3-NEXT:    ret i32 0
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
// CHECK3-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
// CHECK3-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK3-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK3-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK3-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
// CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK3-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK3-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK3-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
// CHECK3-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
// CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK3:       omp.dispatch.cond:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK3:       omp.dispatch.body:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23:![0-9]+]]
// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
// CHECK3-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK3-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
// CHECK3-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK3-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
// CHECK3-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
// CHECK3-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
// CHECK3-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK3-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK3-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
// CHECK3-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
// CHECK3-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
// CHECK3-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
// CHECK3-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
// CHECK3-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
// CHECK3-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
// CHECK3-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK3:       omp.dispatch.inc:
// CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
// CHECK3-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
// CHECK3-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK3:       omp.dispatch.end:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z3bari
// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
// CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
// CHECK3-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
// CHECK3-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
// CHECK3-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
// CHECK3-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
// CHECK3-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
// CHECK3-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
// CHECK3-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
// CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i32 40, i1 false)
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
// CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
// CHECK3-NEXT:    store double* [[A]], double** [[TMP14]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK3-NEXT:    store i32 2, i32* [[TMP22]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
// CHECK3-NEXT:    store i32 2, i32* [[TMP24]], align 4
// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP25]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP30]], align 4
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
// CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
// CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
// CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK3-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK3-NEXT:    store i8* null, i8** [[TMP36]], align 4
// CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP40]], align 4
// CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK3-NEXT:    store i32 5, i32* [[TMP41]], align 4
// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP37]], i8*** [[TMP42]], align 4
// CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP38]], i8*** [[TMP43]], align 4
// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK3-NEXT:    store i64* [[TMP39]], i64** [[TMP44]], align 4
// CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i64** [[TMP45]], align 4
// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP46]], align 4
// CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP47]], align 4
// CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK3-NEXT:    store i64 10, i64* [[TMP48]], align 8
// CHECK3-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK3-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
// CHECK3-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP51:%.*]] = mul nsw i32 1, [[TMP1]]
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP51]]
// CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK3-NEXT:    [[TMP52:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP52]] to i32
// CHECK3-NEXT:    [[TMP53:%.*]] = load i32, i32* [[B]], align 4
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP53]]
// CHECK3-NEXT:    [[TMP54:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP54]])
// CHECK3-NEXT:    ret i32 [[ADD3]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[AAA:%.*]] = alloca i8, align 1
// CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK3-NEXT:    store i8 0, i8* [[AAA]], align 1
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[N_CASTED]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK3-NEXT:    store i8 [[TMP6]], i8* [[CONV1]], align 1
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP12]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP17]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP25]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 4
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 4
// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK3-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK3-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
// CHECK3-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
// CHECK3-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
// CHECK3-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
// CHECK3-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK3-NEXT:    [[ADD6:%.*]] = add i32 [[TMP40]], 1
// CHECK3-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD6]] to i64
// CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP42]], align 4
// CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK3-NEXT:    store i32 5, i32* [[TMP43]], align 4
// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP34]], i8*** [[TMP44]], align 4
// CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP35]], i8*** [[TMP45]], align 4
// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP46]], align 4
// CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP47]], align 4
// CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP48]], align 4
// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP49]], align 4
// CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK3-NEXT:    store i64 [[TMP41]], i64* [[TMP50]], align 8
// CHECK3-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK3-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
// CHECK3-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP53:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    ret i32 [[TMP53]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP9]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP14]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP19]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
// CHECK3-NEXT:    store i32 1, i32* [[TMP22]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
// CHECK3-NEXT:    store i32 3, i32* [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
// CHECK3-NEXT:    store i8** [[TMP20]], i8*** [[TMP24]], align 4
// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
// CHECK3-NEXT:    store i8** [[TMP21]], i8*** [[TMP25]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
// CHECK3-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64** [[TMP26]], align 4
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
// CHECK3-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i64** [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
// CHECK3-NEXT:    store i8** null, i8*** [[TMP28]], align 4
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
// CHECK3-NEXT:    store i8** null, i8*** [[TMP29]], align 4
// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
// CHECK3-NEXT:    store i64 10, i64* [[TMP30]], align 8
// CHECK3-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
// CHECK3-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
// CHECK3-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    ret i32 [[TMP33]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
// CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..12
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
// CHECK3-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    store double [[ADD4]], double* [[A]], align 4
// CHECK3-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
// CHECK3-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK3-NEXT:    store double [[INC]], double* [[A5]], align 4
// CHECK3-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
// CHECK3-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
// CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK3-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
// CHECK3-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK3-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
// CHECK3-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
// CHECK3-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
// CHECK3-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
// CHECK3-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK3:       omp.precond.then:
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK3-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
// CHECK3-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
// CHECK3-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
// CHECK3-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
// CHECK3-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK3-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
// CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
// CHECK3-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
// CHECK3-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
// CHECK3-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
// CHECK3-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
// CHECK3-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
// CHECK3-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK3-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
// CHECK3-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK3:       omp.precond.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
// CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..18
// CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK3-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK3-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK3-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK3-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR4]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
// CHECK9-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
// CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
// CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK9-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK9-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
// CHECK9-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK9-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK9-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK9-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK9-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
// CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK9-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
// CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
// CHECK9-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
// CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
// CHECK9-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
// CHECK9-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK9-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
// CHECK9-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK9-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK9-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK9-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK9-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
// CHECK9-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK9-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK9-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK9-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK9-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
// CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK9:       omp.dispatch.cond:
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK9:       omp.dispatch.body:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13:![0-9]+]]
// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK9-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
// CHECK9-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK9-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
// CHECK9-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
// CHECK9-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
// CHECK9-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK9-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
// CHECK9-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
// CHECK9-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
// CHECK9-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK9-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
// CHECK9-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
// CHECK9-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK9-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
// CHECK9-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
// CHECK9-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
// CHECK9-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
// CHECK9-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK9-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
// CHECK9-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
// CHECK9-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
// CHECK9-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
// CHECK9-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK9:       omp.dispatch.inc:
// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
// CHECK9-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
// CHECK9-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK9:       omp.dispatch.end:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
// CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
// CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
// CHECK9-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK9-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
// CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
// CHECK9-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK9-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
// CHECK9-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I8:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
// CHECK9-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
// CHECK9-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
// CHECK9-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
// CHECK9-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
// CHECK9-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
// CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK9:       omp.precond.then:
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
// CHECK9-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
// CHECK9-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
// CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
// CHECK9-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
// CHECK9-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK9-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
// CHECK9-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
// CHECK9-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
// CHECK9-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
// CHECK9-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
// CHECK9-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
// CHECK9-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
// CHECK9-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
// CHECK9-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
// CHECK9-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK9-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
// CHECK9-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK9:       omp.precond.end:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
// CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
// CHECK9-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK9-NEXT:    store double [[ADD5]], double* [[A]], align 8
// CHECK9-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
// CHECK9-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK9-NEXT:    store double [[INC]], double* [[A6]], align 8
// CHECK9-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
// CHECK9-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
// CHECK9-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK9-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK9-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
// CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK9-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK9:       cond.true:
// CHECK9-NEXT:    br label [[COND_END:%.*]]
// CHECK9:       cond.false:
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
// CHECK9-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
// CHECK9-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
// CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
// CHECK9-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
// CHECK9-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK9-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK9:       omp.body.continue:
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK9-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK9:       omp.inner.for.end:
// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK9:       omp.loop.exit:
// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK9-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
// CHECK11-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
// CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
// CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK11-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK11-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
// CHECK11-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK11-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
// CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK11-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK11-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK11-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
// CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK11-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK11-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
// CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK11-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK11-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK11-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
// CHECK11-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK11-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK11-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK11-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK11-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK11-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK11-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK11-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK11-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
// CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
// CHECK11:       omp.dispatch.cond:
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
// CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
// CHECK11:       omp.dispatch.body:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14:![0-9]+]]
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
// CHECK11-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
// CHECK11-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
// CHECK11-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK11-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
// CHECK11-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK11-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
// CHECK11-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
// CHECK11-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
// CHECK11-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK11-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
// CHECK11-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK11-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
// CHECK11-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
// CHECK11-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
// CHECK11-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
// CHECK11-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK11-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
// CHECK11-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
// CHECK11-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
// CHECK11-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
// CHECK11-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
// CHECK11:       omp.dispatch.inc:
// CHECK11-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
// CHECK11-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
// CHECK11-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[OMP_DISPATCH_COND]]
// CHECK11:       omp.dispatch.end:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
// CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK11-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
// CHECK11-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK11-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
// CHECK11-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
// CHECK11-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
// CHECK11-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
// CHECK11-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
// CHECK11:       omp.precond.then:
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK11-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK11-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
// CHECK11-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
// CHECK11-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
// CHECK11-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
// CHECK11-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK11-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
// CHECK11-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
// CHECK11-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
// CHECK11-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
// CHECK11-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
// CHECK11-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
// CHECK11-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
// CHECK11-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
// CHECK11-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK11-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
// CHECK11-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
// CHECK11:       omp.precond.end:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
// CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK11-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
// CHECK11-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK11-NEXT:    store double [[ADD4]], double* [[A]], align 4
// CHECK11-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
// CHECK11-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK11-NEXT:    store double [[INC]], double* [[A5]], align 4
// CHECK11-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
// CHECK11-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
// CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK11-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
// CHECK11-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
// CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK11-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK11:       cond.true:
// CHECK11-NEXT:    br label [[COND_END:%.*]]
// CHECK11:       cond.false:
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    br label [[COND_END]]
// CHECK11:       cond.end:
// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK11-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
// CHECK11-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
// CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK11-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK11-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
// CHECK11-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK11-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK11:       omp.loop.exit:
// CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
// CHECK11-NEXT:    ret void
//