// 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 -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -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 -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 %s --check-prefix=CHECK2
// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -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 -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 %s --check-prefix=CHECK4

// RUN: %clang_cc1 -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 -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 -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 -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 -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 -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 -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 -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 %s --check-prefix=CHECK9
// RUN: %clang_cc1 -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 -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 %s --check-prefix=CHECK10
// RUN: %clang_cc1 -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 -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 %s --check-prefix=CHECK11
// RUN: %clang_cc1 -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 -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 %s --check-prefix=CHECK12

// RUN: %clang_cc1 -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 -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 -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 -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 -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 -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 -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 -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 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK17
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK18
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK19
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK20

// RUN: %clang_cc1 -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 -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 -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 -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 -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 -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 -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 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK25
// RUN: %clang_cc1 -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 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK26
// RUN: %clang_cc1 -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 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK27
// RUN: %clang_cc1 -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 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK28

// RUN: %clang_cc1 -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 -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 -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 -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 -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 -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 -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 -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 6 that actually will generate offloading
// code and 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 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 parallel nowait
  {
  }

  #pragma omp target parallel if(target: 0)
  {
    a += 1;
  }


  #pragma omp target parallel if(target: 1)
  {
    aa += 1;
#pragma omp cancel parallel
  }


  #pragma omp target parallel if(target: n>10)
  {
    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 parallel if(target: n>20)
  {
    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 parallel if(target: n>40)
  {
    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 parallel if(target: n>50)
  {
    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 parallel if(target: n>60)
    {
      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 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:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[A_CASTED3:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[AA_CASTED5:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[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:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP7]])
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i64 [[TMP12]]) #[[ATTR4:[0-9]+]]
// CHECK1-NEXT:    [[TMP13:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP13]], i16* [[CONV2]], align 2
// CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
// CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP16]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK1-NEXT:    br i1 [[TMP23]], 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_l110(i64 [[TMP14]]) #[[ATTR4]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[A_CASTED3]] to i32*
// CHECK1-NEXT:    store i32 [[TMP24]], i32* [[CONV4]], align 4
// CHECK1-NEXT:    [[TMP25:%.*]] = load i64, i64* [[A_CASTED3]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = load i16, i16* [[AA]], align 2
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED5]] to i16*
// CHECK1-NEXT:    store i16 [[TMP26]], i16* [[CONV6]], align 2
// CHECK1-NEXT:    [[TMP27:%.*]] = load i64, i64* [[AA_CASTED5]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP28]], 10
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
// CHECK1-NEXT:    store i64 [[TMP25]], i64* [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
// CHECK1-NEXT:    store i64 [[TMP25]], i64* [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
// CHECK1-NEXT:    store i64 [[TMP27]], i64* [[TMP35]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK1-NEXT:    store i64 [[TMP27]], i64* [[TMP37]], align 8
// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK1-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK1-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED10:%.*]], label [[OMP_OFFLOAD_CONT11:%.*]]
// CHECK1:       omp_offload.failed10:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR4]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT11]]
// CHECK1:       omp_offload.cont11:
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR4]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
// CHECK1-NEXT:    store i32 [[TMP43]], i32* [[CONV13]], align 4
// CHECK1-NEXT:    [[TMP44:%.*]] = load i64, i64* [[A_CASTED12]], align 8
// CHECK1-NEXT:    [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CMP14:%.*]] = icmp sgt i32 [[TMP45]], 20
// CHECK1-NEXT:    br i1 [[CMP14]], label [[OMP_IF_THEN15:%.*]], label [[OMP_IF_ELSE21:%.*]]
// CHECK1:       omp_if.then15:
// CHECK1-NEXT:    [[TMP46:%.*]] = mul nuw i64 [[TMP2]], 4
// CHECK1-NEXT:    [[TMP47:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK1-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], 8
// CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64*
// CHECK1-NEXT:    store i64 [[TMP44]], i64* [[TMP50]], align 8
// CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
// CHECK1-NEXT:    store i64 [[TMP44]], i64* [[TMP52]], align 8
// CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK1-NEXT:    store i64 4, i64* [[TMP53]], align 8
// CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
// CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 8
// CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
// CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 8
// CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK1-NEXT:    store i64 40, i64* [[TMP59]], align 8
// CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP62]], align 8
// CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP64]], align 8
// CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK1-NEXT:    store i64 8, i64* [[TMP65]], align 8
// CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP66]], align 8
// CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
// CHECK1-NEXT:    store float* [[VLA]], float** [[TMP68]], align 8
// CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
// CHECK1-NEXT:    store float* [[VLA]], float** [[TMP70]], align 8
// CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK1-NEXT:    store i64 [[TMP46]], i64* [[TMP71]], align 8
// CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 3
// CHECK1-NEXT:    store i8* null, i8** [[TMP72]], align 8
// CHECK1-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
// CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 8
// CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
// CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 8
// CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK1-NEXT:    store i64 400, i64* [[TMP77]], align 8
// CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP78]], align 8
// CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
// CHECK1-NEXT:    store i64 5, i64* [[TMP80]], align 8
// CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
// CHECK1-NEXT:    store i64 5, i64* [[TMP82]], align 8
// CHECK1-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK1-NEXT:    store i64 8, i64* [[TMP83]], align 8
// CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 5
// CHECK1-NEXT:    store i8* null, i8** [[TMP84]], align 8
// CHECK1-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP86]], align 8
// CHECK1-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP88]], align 8
// CHECK1-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK1-NEXT:    store i64 8, i64* [[TMP89]], align 8
// CHECK1-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 6
// CHECK1-NEXT:    store i8* null, i8** [[TMP90]], align 8
// CHECK1-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
// CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 8
// CHECK1-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
// CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 8
// CHECK1-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK1-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 8
// CHECK1-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 7
// CHECK1-NEXT:    store i8* null, i8** [[TMP96]], align 8
// CHECK1-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
// CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 8
// CHECK1-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
// CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 8
// CHECK1-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
// CHECK1-NEXT:    store i64 16, i64* [[TMP101]], align 8
// CHECK1-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 8
// CHECK1-NEXT:    store i8* null, i8** [[TMP102]], align 8
// CHECK1-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK1-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
// CHECK1-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
// CHECK1:       omp_offload.failed19:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
// CHECK1:       omp_offload.cont20:
// CHECK1-NEXT:    br label [[OMP_IF_END22:%.*]]
// CHECK1:       omp_if.else21:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK1-NEXT:    br label [[OMP_IF_END22]]
// CHECK1:       omp_if.end22:
// CHECK1-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
// CHECK1-NEXT:    ret i32 [[TMP108]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK1-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK1-SAME: (i32 signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[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:    [[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:%.*]] = 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* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    store i8* [[TMP9]], 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:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
// CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK1:       omp_offload.failed.i:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR4]]
// 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_l104
// CHECK1-SAME: (i64 [[A:%.*]]) #[[ATTR3]] {
// 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 8
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK1-SAME: (i64 [[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 8
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK1-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK1-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK1-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK1:       .cancel.exit:
// CHECK1-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK1-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK1:       .cancel.continue:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK1-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// 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 8
// 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 8
// 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_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK1-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK1-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[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:    [[A_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:    [[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:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.omp_outlined..7 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]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR3]] {
// 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:    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:    [[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:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
// CHECK1-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
// CHECK1-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
// CHECK1-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK1-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
// CHECK1-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] 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* [[ARRAYIDX8]], align 4
// CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
// CHECK1-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK1-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
// CHECK1-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
// CHECK1-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
// CHECK1-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
// CHECK1-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
// CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK1-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
// CHECK1-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
// CHECK1-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
// CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
// CHECK1-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
// CHECK1-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z3bari
// CHECK1-SAME: (i32 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 signext i32 @_Z3fooi(i32 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 signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 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 signext i32 @_ZL7fstatici(i32 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 signext i32 @_Z9ftemplateIiET_i(i32 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* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 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:    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:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
// CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
// CHECK1-NEXT:    store double* [[A]], double** [[TMP13]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK1-NEXT:    store i64 8, i64* [[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 i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK1-NEXT:    store i64 4, i64* [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK1-NEXT:    store i64 2, i64* [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
// CHECK1-NEXT:    store i64 2, i64* [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK1-NEXT:    store i64 8, i64* [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], 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_PTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK1-NEXT:    store i64 8, i64* [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// 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 4
// CHECK1-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
// CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
// CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP39]], align 8
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK1-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK1-NEXT:    br i1 [[TMP44]], 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_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
// 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_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP45:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP45]]
// CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK1-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP46]] to i32
// CHECK1-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP47]]
// CHECK1-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK1-NEXT:    ret i32 [[ADD4]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK1-SAME: (i32 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:    [[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 [4 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
// 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* [[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 i8, i8* [[AAA]], align 1
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK1-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
// CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK1-NEXT:    store i8* null, i8** [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK1-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK1-NEXT:    br i1 [[TMP30]], 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_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// 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_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    ret i32 [[TMP31]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK1-SAME: (i32 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:    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:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK1-NEXT:    br i1 [[TMP23]], 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_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// 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_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    ret i32 [[TMP24]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK1-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* 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 8
// 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_call(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    store double [[ADD]], double* [[A]], align 8
// CHECK1-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
// CHECK1-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK1-NEXT:    store double [[INC]], double* [[A4]], align 8
// CHECK1-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
// CHECK1-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
// CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK1-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK1-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* 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:    [[AAA_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:    [[AAA_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 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* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV2:%.*]] = 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 8
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK1-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// 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:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[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* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV2:%.*]] = 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 8
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK1-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK1-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK1-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
// CHECK1-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
// CHECK1-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK1-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* 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 8
// 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 8
// 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_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK1-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK1-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
// CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
// CHECK2-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
// CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[A_CASTED3:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_CASTED5:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
// CHECK2-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
// CHECK2-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
// CHECK2-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
// CHECK2-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
// CHECK2-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK2-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP7]])
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
// CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i64 [[TMP12]]) #[[ATTR4:[0-9]+]]
// CHECK2-NEXT:    [[TMP13:%.*]] = load i16, i16* [[AA]], align 2
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP13]], i16* [[CONV2]], align 2
// CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
// CHECK2-NEXT:    store i64 [[TMP14]], i64* [[TMP16]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK2-NEXT:    store i64 [[TMP14]], i64* [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK2-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK2-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i64 [[TMP14]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[A_CASTED3]] to i32*
// CHECK2-NEXT:    store i32 [[TMP24]], i32* [[CONV4]], align 4
// CHECK2-NEXT:    [[TMP25:%.*]] = load i64, i64* [[A_CASTED3]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = load i16, i16* [[AA]], align 2
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED5]] to i16*
// CHECK2-NEXT:    store i16 [[TMP26]], i16* [[CONV6]], align 2
// CHECK2-NEXT:    [[TMP27:%.*]] = load i64, i64* [[AA_CASTED5]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP28]], 10
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
// CHECK2-NEXT:    store i64 [[TMP25]], i64* [[TMP30]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
// CHECK2-NEXT:    store i64 [[TMP25]], i64* [[TMP32]], align 8
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
// CHECK2-NEXT:    store i64 [[TMP27]], i64* [[TMP35]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK2-NEXT:    store i64 [[TMP27]], i64* [[TMP37]], align 8
// CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK2-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK2-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED10:%.*]], label [[OMP_OFFLOAD_CONT11:%.*]]
// CHECK2:       omp_offload.failed10:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT11]]
// CHECK2:       omp_offload.cont11:
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
// CHECK2-NEXT:    store i32 [[TMP43]], i32* [[CONV13]], align 4
// CHECK2-NEXT:    [[TMP44:%.*]] = load i64, i64* [[A_CASTED12]], align 8
// CHECK2-NEXT:    [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CMP14:%.*]] = icmp sgt i32 [[TMP45]], 20
// CHECK2-NEXT:    br i1 [[CMP14]], label [[OMP_IF_THEN15:%.*]], label [[OMP_IF_ELSE21:%.*]]
// CHECK2:       omp_if.then15:
// CHECK2-NEXT:    [[TMP46:%.*]] = mul nuw i64 [[TMP2]], 4
// CHECK2-NEXT:    [[TMP47:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK2-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], 8
// CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64*
// CHECK2-NEXT:    store i64 [[TMP44]], i64* [[TMP50]], align 8
// CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
// CHECK2-NEXT:    store i64 [[TMP44]], i64* [[TMP52]], align 8
// CHECK2-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK2-NEXT:    store i64 4, i64* [[TMP53]], align 8
// CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK2-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
// CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 8
// CHECK2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
// CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 8
// CHECK2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK2-NEXT:    store i64 40, i64* [[TMP59]], align 8
// CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK2-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
// CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP62]], align 8
// CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
// CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP64]], align 8
// CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK2-NEXT:    store i64 8, i64* [[TMP65]], align 8
// CHECK2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP66]], align 8
// CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
// CHECK2-NEXT:    store float* [[VLA]], float** [[TMP68]], align 8
// CHECK2-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
// CHECK2-NEXT:    store float* [[VLA]], float** [[TMP70]], align 8
// CHECK2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK2-NEXT:    store i64 [[TMP46]], i64* [[TMP71]], align 8
// CHECK2-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 3
// CHECK2-NEXT:    store i8* null, i8** [[TMP72]], align 8
// CHECK2-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
// CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 8
// CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
// CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 8
// CHECK2-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK2-NEXT:    store i64 400, i64* [[TMP77]], align 8
// CHECK2-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 4
// CHECK2-NEXT:    store i8* null, i8** [[TMP78]], align 8
// CHECK2-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 5
// CHECK2-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
// CHECK2-NEXT:    store i64 5, i64* [[TMP80]], align 8
// CHECK2-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 5
// CHECK2-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
// CHECK2-NEXT:    store i64 5, i64* [[TMP82]], align 8
// CHECK2-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK2-NEXT:    store i64 8, i64* [[TMP83]], align 8
// CHECK2-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 5
// CHECK2-NEXT:    store i8* null, i8** [[TMP84]], align 8
// CHECK2-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 6
// CHECK2-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP86]], align 8
// CHECK2-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 6
// CHECK2-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP88]], align 8
// CHECK2-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK2-NEXT:    store i64 8, i64* [[TMP89]], align 8
// CHECK2-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 6
// CHECK2-NEXT:    store i8* null, i8** [[TMP90]], align 8
// CHECK2-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 7
// CHECK2-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
// CHECK2-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 8
// CHECK2-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 7
// CHECK2-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
// CHECK2-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 8
// CHECK2-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK2-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 8
// CHECK2-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 7
// CHECK2-NEXT:    store i8* null, i8** [[TMP96]], align 8
// CHECK2-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 8
// CHECK2-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
// CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 8
// CHECK2-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 8
// CHECK2-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
// CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 8
// CHECK2-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
// CHECK2-NEXT:    store i64 16, i64* [[TMP101]], align 8
// CHECK2-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 8
// CHECK2-NEXT:    store i8* null, i8** [[TMP102]], align 8
// CHECK2-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK2-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
// CHECK2-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
// CHECK2:       omp_offload.failed19:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
// CHECK2:       omp_offload.cont20:
// CHECK2-NEXT:    br label [[OMP_IF_END22:%.*]]
// CHECK2:       omp_if.else21:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_IF_END22]]
// CHECK2:       omp_if.end22:
// CHECK2-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
// CHECK2-NEXT:    ret i32 [[TMP108]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK2-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK2-SAME: (i32 signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
// CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
// CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
// CHECK2-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
// CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
// CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
// CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
// CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
// CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
// CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
// CHECK2-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
// CHECK2-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
// CHECK2-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
// CHECK2-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK2-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
// CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK2-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK2:       omp_offload.failed.i:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR4]]
// CHECK2-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
// CHECK2:       .omp_outlined..1.exit:
// CHECK2-NEXT:    ret i32 0
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
// CHECK2-SAME: (i64 [[A:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK2-SAME: (i64 [[AA:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK2-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK2-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK2-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK2-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK2:       .cancel.exit:
// CHECK2-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK2-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK2:       .cancel.continue:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK2-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
// CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK2-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK2-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK2-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.omp_outlined..7 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]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK2-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
// CHECK2-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
// CHECK2-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
// CHECK2-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK2-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
// CHECK2-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
// CHECK2-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
// CHECK2-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
// CHECK2-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
// CHECK2-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
// CHECK2-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK2-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK2-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
// CHECK2-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
// CHECK2-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
// CHECK2-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK2-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
// CHECK2-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
// CHECK2-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK2-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
// CHECK2-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
// CHECK2-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
// CHECK2-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
// CHECK2-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
// CHECK2-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z3bari
// CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
// CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_Z3fooi(i32 signext [[TMP0]])
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP2]])
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
// CHECK2-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CALL3:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP4]])
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
// CHECK2-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CALL5:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP6]])
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
// CHECK2-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
// CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
// CHECK2-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
// CHECK2-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
// CHECK2-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
// CHECK2-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
// CHECK2-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
// CHECK2-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
// CHECK2-NEXT:    store double* [[A]], double** [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK2-NEXT:    store i64 8, i64* [[TMP14]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK2-NEXT:    store i64 4, i64* [[TMP20]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK2-NEXT:    store i64 2, i64* [[TMP23]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
// CHECK2-NEXT:    store i64 2, i64* [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK2-NEXT:    store i64 8, i64* [[TMP26]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
// CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP31]], align 8
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK2-NEXT:    store i64 8, i64* [[TMP32]], align 8
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK2-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
// CHECK2-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
// CHECK2-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
// CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK2-NEXT:    store i8* null, i8** [[TMP39]], align 8
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK2-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK2-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    [[TMP45:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP45]]
// CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK2-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK2-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP46]] to i32
// CHECK2-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
// CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP47]]
// CHECK2-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK2-NEXT:    ret i32 [[ADD4]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[AAA:%.*]] = alloca i8, align 1
// CHECK2-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK2-NEXT:    store i8 0, i8* [[AAA]], align 1
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK2-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
// CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
// CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK2-NEXT:    store i8* null, i8** [[TMP26]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK2-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK2-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    ret i32 [[TMP31]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
// CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP14]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
// CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
// CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK2-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK2-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    ret i32 [[TMP24]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK2-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK2-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    store double [[ADD]], double* [[A]], align 8
// CHECK2-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
// CHECK2-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK2-NEXT:    store double [[INC]], double* [[A4]], align 8
// CHECK2-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
// CHECK2-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
// CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK2-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK2-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK2-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
// CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK2-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK2-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK2-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
// CHECK2-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
// CHECK2-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK2-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK2-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK2-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK2-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK2-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK3-SAME: (i32 [[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:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[AA_CASTED3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
// CHECK3-NEXT:    [[A_CASTED10:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[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:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP5]])
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i32 [[TMP10]]) #[[ATTR4:[0-9]+]]
// CHECK3-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
// CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP18]], i8** [[TMP19]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK3-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK3-NEXT:    br i1 [[TMP21]], 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_l110(i32 [[TMP12]]) #[[ATTR4]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP22]], i32* [[A_CASTED2]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A_CASTED2]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = load i16, i16* [[AA]], align 2
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i32* [[AA_CASTED3]] to i16*
// CHECK3-NEXT:    store i16 [[TMP24]], i16* [[CONV4]], align 2
// CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[AA_CASTED3]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP26]], 10
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
// CHECK3-NEXT:    store i32 [[TMP23]], i32* [[TMP28]], align 4
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK3-NEXT:    store i32 [[TMP23]], i32* [[TMP30]], align 4
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32*
// CHECK3-NEXT:    store i32 [[TMP25]], i32* [[TMP33]], align 4
// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
// CHECK3-NEXT:    store i32 [[TMP25]], i32* [[TMP35]], align 4
// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP36]], align 4
// CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP37]], i8** [[TMP38]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK3-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK3-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK3:       omp_offload.failed8:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR4]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK3:       omp_offload.cont9:
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR4]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP41]], i32* [[A_CASTED10]], align 4
// CHECK3-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A_CASTED10]], align 4
// CHECK3-NEXT:    [[TMP43:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], 20
// CHECK3-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN12:%.*]], label [[OMP_IF_ELSE18:%.*]]
// CHECK3:       omp_if.then12:
// CHECK3-NEXT:    [[TMP44:%.*]] = mul nuw i32 [[TMP1]], 4
// CHECK3-NEXT:    [[TMP45:%.*]] = sext i32 [[TMP44]] to i64
// CHECK3-NEXT:    [[TMP46:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK3-NEXT:    [[TMP47:%.*]] = mul nuw i32 [[TMP46]], 8
// CHECK3-NEXT:    [[TMP48:%.*]] = sext i32 [[TMP47]] to i64
// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
// CHECK3-NEXT:    store i32 [[TMP42]], i32* [[TMP50]], align 4
// CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK3-NEXT:    store i32 [[TMP42]], i32* [[TMP52]], align 4
// CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK3-NEXT:    store i64 4, i64* [[TMP53]], align 4
// CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK3-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
// CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 4
// CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
// CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 4
// CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK3-NEXT:    store i64 40, i64* [[TMP59]], align 4
// CHECK3-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP60]], align 4
// CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP62]], align 4
// CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP64]], align 4
// CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK3-NEXT:    store i64 4, i64* [[TMP65]], align 4
// CHECK3-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP66]], align 4
// CHECK3-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
// CHECK3-NEXT:    store float* [[VLA]], float** [[TMP68]], align 4
// CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
// CHECK3-NEXT:    store float* [[VLA]], float** [[TMP70]], align 4
// CHECK3-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK3-NEXT:    store i64 [[TMP45]], i64* [[TMP71]], align 4
// CHECK3-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP72]], align 4
// CHECK3-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
// CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 4
// CHECK3-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
// CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 4
// CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK3-NEXT:    store i64 400, i64* [[TMP77]], align 4
// CHECK3-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 4
// CHECK3-NEXT:    store i8* null, i8** [[TMP78]], align 4
// CHECK3-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
// CHECK3-NEXT:    store i32 5, i32* [[TMP80]], align 4
// CHECK3-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
// CHECK3-NEXT:    store i32 5, i32* [[TMP82]], align 4
// CHECK3-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK3-NEXT:    store i64 4, i64* [[TMP83]], align 4
// CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 5
// CHECK3-NEXT:    store i8* null, i8** [[TMP84]], align 4
// CHECK3-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP86]], align 4
// CHECK3-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP88]], align 4
// CHECK3-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK3-NEXT:    store i64 4, i64* [[TMP89]], align 4
// CHECK3-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 6
// CHECK3-NEXT:    store i8* null, i8** [[TMP90]], align 4
// CHECK3-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
// CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 4
// CHECK3-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
// CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 4
// CHECK3-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK3-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 4
// CHECK3-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 7
// CHECK3-NEXT:    store i8* null, i8** [[TMP96]], align 4
// CHECK3-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
// CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 4
// CHECK3-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
// CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 4
// CHECK3-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
// CHECK3-NEXT:    store i64 12, i64* [[TMP101]], align 4
// CHECK3-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 8
// CHECK3-NEXT:    store i8* null, i8** [[TMP102]], align 4
// CHECK3-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK3-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
// CHECK3-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK3:       omp_offload.failed16:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK3:       omp_offload.cont17:
// CHECK3-NEXT:    br label [[OMP_IF_END19:%.*]]
// CHECK3:       omp_if.else18:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK3-NEXT:    br label [[OMP_IF_END19]]
// CHECK3:       omp_if.end19:
// CHECK3-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
// CHECK3-NEXT:    ret i32 [[TMP108]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK3-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = 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:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK3-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[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:    [[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:%.*]] = 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* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    store i8* [[TMP9]], 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:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
// CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK3-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK3:       omp_offload.failed.i:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR4]]
// 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_l104
// CHECK3-SAME: (i32 [[A:%.*]]) #[[ATTR3]] {
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK3-SAME: (i32 [[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 4
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK3-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK3-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK3-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK3-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK3:       .cancel.exit:
// CHECK3-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK3-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK3:       .cancel.continue:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK3-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// 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 4
// 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_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK3-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK3-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK3-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[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:    [[A_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:    [[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:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.omp_outlined..7 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]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR3]] {
// 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:    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:    [[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:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
// CHECK3-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK3-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
// CHECK3-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK3-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
// CHECK3-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
// CHECK3-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
// CHECK3-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
// CHECK3-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
// CHECK3-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK3-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
// CHECK3-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
// CHECK3-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK3-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
// CHECK3-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
// CHECK3-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
// CHECK3-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
// CHECK3-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
// CHECK3-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK3-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
// CHECK3-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
// CHECK3-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
// CHECK3-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
// CHECK3-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
// CHECK3-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z3bari
// CHECK3-SAME: (i32 [[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 i32 @_Z3fooi(i32 [[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 i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[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 i32 @_ZL7fstatici(i32 [[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 i32 @_Z9ftemplateIiET_i(i32 [[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* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[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:    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:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
// CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
// CHECK3-NEXT:    store double* [[A]], double** [[TMP13]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK3-NEXT:    store i64 8, i64* [[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 i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK3-NEXT:    store i64 4, i64* [[TMP20]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK3-NEXT:    store i32 2, i32* [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
// CHECK3-NEXT:    store i32 2, i32* [[TMP25]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK3-NEXT:    store i64 4, i64* [[TMP26]], align 4
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], 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_PTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP31]], align 4
// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK3-NEXT:    store i64 4, i64* [[TMP32]], align 4
// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// 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 4
// CHECK3-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
// CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
// CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK3-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 4
// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK3-NEXT:    store i8* null, i8** [[TMP39]], align 4
// CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK3-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK3-NEXT:    br i1 [[TMP44]], 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_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
// 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_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP45:%.*]] = mul nsw i32 1, [[TMP1]]
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP45]]
// CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK3-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP46]] to i32
// CHECK3-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP47]]
// CHECK3-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK3-NEXT:    ret i32 [[ADD3]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK3-SAME: (i32 [[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:    [[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 [4 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], 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* [[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 i8, i8* [[AAA]], align 1
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK3-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
// CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP26]], align 4
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK3-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK3-NEXT:    br i1 [[TMP30]], 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_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// 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_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    ret i32 [[TMP31]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK3-SAME: (i32 [[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:    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:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK3-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK3-NEXT:    br i1 [[TMP23]], 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_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// 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_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    ret i32 [[TMP24]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK3-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* 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_call(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    store double [[ADD]], double* [[A]], align 4
// CHECK3-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
// CHECK3-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK3-NEXT:    store double [[INC]], double* [[A3]], align 4
// CHECK3-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
// CHECK3-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
// CHECK3-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK3-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK3-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* 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:    [[AAA_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:    [[AAA_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 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* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK3-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// 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:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x 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 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:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] 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 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK3-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
// CHECK3-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
// CHECK3-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK3-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK3-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* 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 4
// 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_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// 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:    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:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK3-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK3-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK3-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
// CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
// CHECK4-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
// CHECK4-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK4-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_CASTED3:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
// CHECK4-NEXT:    [[A_CASTED10:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK4-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK4-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
// CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP5]])
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i32 [[TMP10]]) #[[ATTR4:[0-9]+]]
// CHECK4-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
// CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK4-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
// CHECK4-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP18]], i8** [[TMP19]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK4-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK4-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i32 [[TMP12]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    store i32 [[TMP22]], i32* [[A_CASTED2]], align 4
// CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A_CASTED2]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = load i16, i16* [[AA]], align 2
// CHECK4-NEXT:    [[CONV4:%.*]] = bitcast i32* [[AA_CASTED3]] to i16*
// CHECK4-NEXT:    store i16 [[TMP24]], i16* [[CONV4]], align 2
// CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[AA_CASTED3]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP26]], 10
// CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
// CHECK4-NEXT:    store i32 [[TMP23]], i32* [[TMP28]], align 4
// CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK4-NEXT:    store i32 [[TMP23]], i32* [[TMP30]], align 4
// CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32*
// CHECK4-NEXT:    store i32 [[TMP25]], i32* [[TMP33]], align 4
// CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
// CHECK4-NEXT:    store i32 [[TMP25]], i32* [[TMP35]], align 4
// CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP36]], align 4
// CHECK4-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP37]], i8** [[TMP38]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK4-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK4-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK4:       omp_offload.failed8:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK4:       omp_offload.cont9:
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    store i32 [[TMP41]], i32* [[A_CASTED10]], align 4
// CHECK4-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A_CASTED10]], align 4
// CHECK4-NEXT:    [[TMP43:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], 20
// CHECK4-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN12:%.*]], label [[OMP_IF_ELSE18:%.*]]
// CHECK4:       omp_if.then12:
// CHECK4-NEXT:    [[TMP44:%.*]] = mul nuw i32 [[TMP1]], 4
// CHECK4-NEXT:    [[TMP45:%.*]] = sext i32 [[TMP44]] to i64
// CHECK4-NEXT:    [[TMP46:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK4-NEXT:    [[TMP47:%.*]] = mul nuw i32 [[TMP46]], 8
// CHECK4-NEXT:    [[TMP48:%.*]] = sext i32 [[TMP47]] to i64
// CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
// CHECK4-NEXT:    store i32 [[TMP42]], i32* [[TMP50]], align 4
// CHECK4-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK4-NEXT:    store i32 [[TMP42]], i32* [[TMP52]], align 4
// CHECK4-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK4-NEXT:    store i64 4, i64* [[TMP53]], align 4
// CHECK4-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK4-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
// CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 4
// CHECK4-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
// CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 4
// CHECK4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK4-NEXT:    store i64 40, i64* [[TMP59]], align 4
// CHECK4-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP60]], align 4
// CHECK4-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP62]], align 4
// CHECK4-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP64]], align 4
// CHECK4-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK4-NEXT:    store i64 4, i64* [[TMP65]], align 4
// CHECK4-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP66]], align 4
// CHECK4-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
// CHECK4-NEXT:    store float* [[VLA]], float** [[TMP68]], align 4
// CHECK4-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
// CHECK4-NEXT:    store float* [[VLA]], float** [[TMP70]], align 4
// CHECK4-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK4-NEXT:    store i64 [[TMP45]], i64* [[TMP71]], align 4
// CHECK4-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 3
// CHECK4-NEXT:    store i8* null, i8** [[TMP72]], align 4
// CHECK4-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
// CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 4
// CHECK4-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
// CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 4
// CHECK4-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK4-NEXT:    store i64 400, i64* [[TMP77]], align 4
// CHECK4-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 4
// CHECK4-NEXT:    store i8* null, i8** [[TMP78]], align 4
// CHECK4-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
// CHECK4-NEXT:    store i32 5, i32* [[TMP80]], align 4
// CHECK4-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
// CHECK4-NEXT:    store i32 5, i32* [[TMP82]], align 4
// CHECK4-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK4-NEXT:    store i64 4, i64* [[TMP83]], align 4
// CHECK4-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 5
// CHECK4-NEXT:    store i8* null, i8** [[TMP84]], align 4
// CHECK4-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP86]], align 4
// CHECK4-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP88]], align 4
// CHECK4-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK4-NEXT:    store i64 4, i64* [[TMP89]], align 4
// CHECK4-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 6
// CHECK4-NEXT:    store i8* null, i8** [[TMP90]], align 4
// CHECK4-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 7
// CHECK4-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
// CHECK4-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 4
// CHECK4-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 7
// CHECK4-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
// CHECK4-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 4
// CHECK4-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK4-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 4
// CHECK4-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 7
// CHECK4-NEXT:    store i8* null, i8** [[TMP96]], align 4
// CHECK4-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 8
// CHECK4-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
// CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 4
// CHECK4-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 8
// CHECK4-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
// CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 4
// CHECK4-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
// CHECK4-NEXT:    store i64 12, i64* [[TMP101]], align 4
// CHECK4-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 8
// CHECK4-NEXT:    store i8* null, i8** [[TMP102]], align 4
// CHECK4-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK4-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
// CHECK4-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK4:       omp_offload.failed16:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK4:       omp_offload.cont17:
// CHECK4-NEXT:    br label [[OMP_IF_END19:%.*]]
// CHECK4:       omp_if.else18:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_IF_END19]]
// CHECK4:       omp_if.end19:
// CHECK4-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
// CHECK4-NEXT:    ret i32 [[TMP108]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK4-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK4-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
// CHECK4-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
// CHECK4-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
// CHECK4-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
// CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
// CHECK4-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
// CHECK4-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
// CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
// CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
// CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
// CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
// CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
// CHECK4-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
// CHECK4-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
// CHECK4-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
// CHECK4-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
// CHECK4-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK4-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
// CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK4-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK4:       omp_offload.failed.i:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR4]]
// CHECK4-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
// CHECK4:       .omp_outlined..1.exit:
// CHECK4-NEXT:    ret i32 0
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
// CHECK4-SAME: (i32 [[A:%.*]]) #[[ATTR3]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]]) #[[ATTR3]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK4-SAME: (i32 [[AA:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR3]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK4-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK4-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK4-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK4-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK4:       .cancel.exit:
// CHECK4-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK4-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK4:       .cancel.continue:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK4-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR3]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK4-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK4-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK4-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK4-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK4-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK4-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.omp_outlined..7 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]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR3]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK4-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK4-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK4-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
// CHECK4-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK4-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
// CHECK4-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK4-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
// CHECK4-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
// CHECK4-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
// CHECK4-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
// CHECK4-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
// CHECK4-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK4-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
// CHECK4-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK4-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
// CHECK4-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK4-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
// CHECK4-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
// CHECK4-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
// CHECK4-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK4-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
// CHECK4-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
// CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK4-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
// CHECK4-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
// CHECK4-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
// CHECK4-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
// CHECK4-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
// CHECK4-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z3bari
// CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooi(i32 [[TMP0]])
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CALL1:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP2]])
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
// CHECK4-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CALL3:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP4]])
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
// CHECK4-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CALL5:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP6]])
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
// CHECK4-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
// CHECK4-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
// CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
// CHECK4-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
// CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
// CHECK4-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
// CHECK4-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
// CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
// CHECK4-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 4
// CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
// CHECK4-NEXT:    store double* [[A]], double** [[TMP13]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK4-NEXT:    store i64 8, i64* [[TMP14]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP15]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK4-NEXT:    store i64 4, i64* [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK4-NEXT:    store i32 2, i32* [[TMP23]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
// CHECK4-NEXT:    store i32 2, i32* [[TMP25]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK4-NEXT:    store i64 4, i64* [[TMP26]], align 4
// CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP27]], align 4
// CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
// CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP31]], align 4
// CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK4-NEXT:    store i64 4, i64* [[TMP32]], align 4
// CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK4-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
// CHECK4-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
// CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
// CHECK4-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
// CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK4-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 4
// CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK4-NEXT:    store i8* null, i8** [[TMP39]], align 4
// CHECK4-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK4-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK4-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    [[TMP45:%.*]] = mul nsw i32 1, [[TMP1]]
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP45]]
// CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK4-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP46]] to i32
// CHECK4-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
// CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP47]]
// CHECK4-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK4-NEXT:    ret i32 [[ADD3]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[AAA:%.*]] = alloca i8, align 1
// CHECK4-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK4-NEXT:    store i8 0, i8* [[AAA]], align 1
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK4-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
// CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
// CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
// CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
// CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK4-NEXT:    store i8* null, i8** [[TMP26]], align 4
// CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK4-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK4-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    ret i32 [[TMP31]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
// CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
// CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP9]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
// CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP14]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
// CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
// CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP19]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK4-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK4-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    ret i32 [[TMP24]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK4-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR3]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK4-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK4-NEXT:    store double [[ADD]], double* [[A]], align 4
// CHECK4-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
// CHECK4-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK4-NEXT:    store double [[INC]], double* [[A3]], align 4
// CHECK4-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
// CHECK4-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
// CHECK4-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK4-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK4-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK4-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK4-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK4-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK4-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
// CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
// CHECK4-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
// CHECK4-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK4-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK4-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
// CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK4-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK4-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK4-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK4-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK4-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// 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:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK9-SAME: (i64 [[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 8
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR1]] {
// 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:    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:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK9-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK9-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK9-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK9-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK9-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK9-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK9:       .cancel.exit:
// CHECK9-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK9-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK9:       .cancel.continue:
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK9-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// 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 8
// 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 8
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR1]] {
// 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:    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:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK9-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK9-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK9-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[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:    [[A_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:    [[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:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
// CHECK9-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR1]] {
// 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:    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:    [[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:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK9-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
// CHECK9-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
// CHECK9-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
// CHECK9-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK9-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
// CHECK9-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] 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* [[ARRAYIDX8]], align 4
// CHECK9-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK9-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
// CHECK9-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
// CHECK9-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK9-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
// CHECK9-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK9-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
// CHECK9-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
// CHECK9-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
// CHECK9-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK9-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
// CHECK9-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
// CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK9-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
// CHECK9-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK9-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
// CHECK9-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
// CHECK9-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
// CHECK9-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
// CHECK9-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK9-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* 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:    [[AAA_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:    [[AAA_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 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* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[CONV2:%.*]] = 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 8
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK9-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK9-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
// CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// 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:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// 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 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* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK9-NEXT:    [[CONV2:%.*]] = 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 8
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK9-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
// CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK9-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK9-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK9-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
// CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
// CHECK9-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
// CHECK9-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK9-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK9-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* 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 8
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
// 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:    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:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK9-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK9-NEXT:    store double [[ADD]], double* [[A]], align 8
// CHECK9-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
// CHECK9-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK9-NEXT:    store double [[INC]], double* [[A4]], align 8
// CHECK9-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
// CHECK9-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
// CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK9-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK9-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* 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 8
// 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 8
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// 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:    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:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK9-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK9-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK9-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
// CHECK9-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK10-SAME: (i64 [[AA:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK10-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK10-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR1]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK10-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK10-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK10-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK10-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK10-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK10-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK10:       .cancel.exit:
// CHECK10-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK10-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK10:       .cancel.continue:
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK10-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
// CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK10-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
// CHECK10-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR1]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK10-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK10-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK10-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK10-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK10-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK10-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK10-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK10-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK10-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK10-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK10-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
// CHECK10-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR1]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK10-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK10-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK10-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK10-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK10-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK10-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK10-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK10-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK10-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
// CHECK10-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
// CHECK10-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
// CHECK10-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK10-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
// CHECK10-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
// CHECK10-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
// CHECK10-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
// CHECK10-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
// CHECK10-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK10-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
// CHECK10-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
// CHECK10-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK10-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
// CHECK10-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK10-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
// CHECK10-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
// CHECK10-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
// CHECK10-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK10-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
// CHECK10-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
// CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK10-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
// CHECK10-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK10-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
// CHECK10-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
// CHECK10-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
// CHECK10-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
// CHECK10-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK10-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK10-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
// CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK10-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK10-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
// CHECK10-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
// CHECK10-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK10-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK10-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK10-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
// CHECK10-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
// CHECK10-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
// CHECK10-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK10-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK10-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK10-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK10-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
// CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK10-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK10-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
// CHECK10-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK10-NEXT:    store double [[ADD]], double* [[A]], align 8
// CHECK10-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK10-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
// CHECK10-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK10-NEXT:    store double [[INC]], double* [[A4]], align 8
// CHECK10-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
// CHECK10-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
// CHECK10-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK10-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK10-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
// CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK10-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK10-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK10-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK10-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
// CHECK10-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = 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:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK11-SAME: (i32 [[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 4
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR1]] {
// 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:    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:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK11-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK11-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK11-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK11-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK11:       .cancel.exit:
// CHECK11-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK11-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK11:       .cancel.continue:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK11-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// 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 4
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR1]] {
// 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:    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:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK11-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK11-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK11-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[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:    [[A_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:    [[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:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR1]] {
// 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:    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:    [[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:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
// CHECK11-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK11-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
// CHECK11-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK11-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
// CHECK11-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
// CHECK11-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
// CHECK11-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
// CHECK11-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
// CHECK11-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK11-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
// CHECK11-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK11-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
// CHECK11-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK11-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
// CHECK11-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
// CHECK11-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
// CHECK11-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK11-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
// CHECK11-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
// CHECK11-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK11-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
// CHECK11-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK11-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
// CHECK11-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
// CHECK11-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
// CHECK11-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
// CHECK11-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK11-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* 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:    [[AAA_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:    [[AAA_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 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* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK11-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK11-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK11-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK11-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// 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:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x 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 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:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK11-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] 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 4
// CHECK11-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK11-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
// CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
// CHECK11-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
// CHECK11-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK11-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK11-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
// 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:    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:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK11-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK11-NEXT:    store double [[ADD]], double* [[A]], align 4
// CHECK11-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
// CHECK11-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK11-NEXT:    store double [[INC]], double* [[A3]], align 4
// CHECK11-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
// CHECK11-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
// CHECK11-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK11-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK11-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* 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 4
// 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_call(%struct.ident_t* @[[GLOB1]], 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 [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// 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:    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:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
// CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK11-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK11-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK11-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK12-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK12-SAME: (i32 [[AA:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK12-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK12-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR1]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK12-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK12-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK12-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK12-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK12-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK12:       .cancel.exit:
// CHECK12-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK12-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK12:       .cancel.continue:
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK12-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK12-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR1]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK12-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK12-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK12-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK12-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK12-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK12-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK12-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK12-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK12-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK12-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR1]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK12-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK12-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK12-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK12-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK12-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK12-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK12-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK12-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK12-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
// CHECK12-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK12-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
// CHECK12-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
// CHECK12-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK12-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
// CHECK12-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
// CHECK12-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
// CHECK12-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
// CHECK12-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
// CHECK12-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK12-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
// CHECK12-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
// CHECK12-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK12-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
// CHECK12-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK12-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
// CHECK12-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
// CHECK12-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
// CHECK12-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK12-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
// CHECK12-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK12-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
// CHECK12-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK12-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
// CHECK12-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK12-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
// CHECK12-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
// CHECK12-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
// CHECK12-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
// CHECK12-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK12-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK12-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK12-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK12-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
// CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK12-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK12-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK12-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
// CHECK12-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
// CHECK12-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
// CHECK12-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK12-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK12-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK12-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK12-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK12-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK12-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK12-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK12-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK12-NEXT:    store double [[ADD]], double* [[A]], align 4
// CHECK12-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK12-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
// CHECK12-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK12-NEXT:    store double [[INC]], double* [[A3]], align 4
// CHECK12-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
// CHECK12-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
// CHECK12-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK12-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK12-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK12-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
// CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK12-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK12-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK12-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK12-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
// CHECK12-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK17-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
// CHECK17-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
// CHECK17-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[A_CASTED3:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_CASTED5:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
// CHECK17-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [9 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [9 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [9 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
// CHECK17-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
// CHECK17-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
// CHECK17-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
// CHECK17-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
// CHECK17-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK17-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP7]])
// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i64 [[TMP12]]) #[[ATTR4:[0-9]+]]
// CHECK17-NEXT:    [[TMP13:%.*]] = load i16, i16* [[AA]], align 2
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP13]], i16* [[CONV2]], align 2
// CHECK17-NEXT:    [[TMP14:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
// CHECK17-NEXT:    store i64 [[TMP14]], i64* [[TMP16]], align 8
// CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK17-NEXT:    store i64 [[TMP14]], i64* [[TMP18]], align 8
// CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK17:       omp_offload.failed:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i64 [[TMP14]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[A_CASTED3]] to i32*
// CHECK17-NEXT:    store i32 [[TMP24]], i32* [[CONV4]], align 4
// CHECK17-NEXT:    [[TMP25:%.*]] = load i64, i64* [[A_CASTED3]], align 8
// CHECK17-NEXT:    [[TMP26:%.*]] = load i16, i16* [[AA]], align 2
// CHECK17-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED5]] to i16*
// CHECK17-NEXT:    store i16 [[TMP26]], i16* [[CONV6]], align 2
// CHECK17-NEXT:    [[TMP27:%.*]] = load i64, i64* [[AA_CASTED5]], align 8
// CHECK17-NEXT:    [[TMP28:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP28]], 10
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK17:       omp_if.then:
// CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
// CHECK17-NEXT:    store i64 [[TMP25]], i64* [[TMP30]], align 8
// CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
// CHECK17-NEXT:    store i64 [[TMP25]], i64* [[TMP32]], align 8
// CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
// CHECK17-NEXT:    store i64 [[TMP27]], i64* [[TMP35]], align 8
// CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK17-NEXT:    store i64 [[TMP27]], i64* [[TMP37]], align 8
// CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK17-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK17-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED10:%.*]], label [[OMP_OFFLOAD_CONT11:%.*]]
// CHECK17:       omp_offload.failed10:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT11]]
// CHECK17:       omp_offload.cont11:
// CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK17:       omp_if.else:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_IF_END]]
// CHECK17:       omp_if.end:
// CHECK17-NEXT:    [[TMP43:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
// CHECK17-NEXT:    store i32 [[TMP43]], i32* [[CONV13]], align 4
// CHECK17-NEXT:    [[TMP44:%.*]] = load i64, i64* [[A_CASTED12]], align 8
// CHECK17-NEXT:    [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CMP14:%.*]] = icmp sgt i32 [[TMP45]], 20
// CHECK17-NEXT:    br i1 [[CMP14]], label [[OMP_IF_THEN15:%.*]], label [[OMP_IF_ELSE21:%.*]]
// CHECK17:       omp_if.then15:
// CHECK17-NEXT:    [[TMP46:%.*]] = mul nuw i64 [[TMP2]], 4
// CHECK17-NEXT:    [[TMP47:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK17-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], 8
// CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64*
// CHECK17-NEXT:    store i64 [[TMP44]], i64* [[TMP50]], align 8
// CHECK17-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
// CHECK17-NEXT:    store i64 [[TMP44]], i64* [[TMP52]], align 8
// CHECK17-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    store i64 4, i64* [[TMP53]], align 8
// CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
// CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 8
// CHECK17-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
// CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 8
// CHECK17-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK17-NEXT:    store i64 40, i64* [[TMP59]], align 8
// CHECK17-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK17-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
// CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP62]], align 8
// CHECK17-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
// CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP64]], align 8
// CHECK17-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK17-NEXT:    store i64 8, i64* [[TMP65]], align 8
// CHECK17-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP66]], align 8
// CHECK17-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
// CHECK17-NEXT:    store float* [[VLA]], float** [[TMP68]], align 8
// CHECK17-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
// CHECK17-NEXT:    store float* [[VLA]], float** [[TMP70]], align 8
// CHECK17-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK17-NEXT:    store i64 [[TMP46]], i64* [[TMP71]], align 8
// CHECK17-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP72]], align 8
// CHECK17-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 4
// CHECK17-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
// CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 8
// CHECK17-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 4
// CHECK17-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
// CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 8
// CHECK17-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK17-NEXT:    store i64 400, i64* [[TMP77]], align 8
// CHECK17-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 4
// CHECK17-NEXT:    store i8* null, i8** [[TMP78]], align 8
// CHECK17-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 5
// CHECK17-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
// CHECK17-NEXT:    store i64 5, i64* [[TMP80]], align 8
// CHECK17-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 5
// CHECK17-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
// CHECK17-NEXT:    store i64 5, i64* [[TMP82]], align 8
// CHECK17-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK17-NEXT:    store i64 8, i64* [[TMP83]], align 8
// CHECK17-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 5
// CHECK17-NEXT:    store i8* null, i8** [[TMP84]], align 8
// CHECK17-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 6
// CHECK17-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP86]], align 8
// CHECK17-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 6
// CHECK17-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP88]], align 8
// CHECK17-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK17-NEXT:    store i64 8, i64* [[TMP89]], align 8
// CHECK17-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 6
// CHECK17-NEXT:    store i8* null, i8** [[TMP90]], align 8
// CHECK17-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 7
// CHECK17-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
// CHECK17-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 8
// CHECK17-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 7
// CHECK17-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
// CHECK17-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 8
// CHECK17-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK17-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 8
// CHECK17-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 7
// CHECK17-NEXT:    store i8* null, i8** [[TMP96]], align 8
// CHECK17-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 8
// CHECK17-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
// CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 8
// CHECK17-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 8
// CHECK17-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
// CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 8
// CHECK17-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
// CHECK17-NEXT:    store i64 16, i64* [[TMP101]], align 8
// CHECK17-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 8
// CHECK17-NEXT:    store i8* null, i8** [[TMP102]], align 8
// CHECK17-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK17-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
// CHECK17-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
// CHECK17:       omp_offload.failed19:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
// CHECK17:       omp_offload.cont20:
// CHECK17-NEXT:    br label [[OMP_IF_END22:%.*]]
// CHECK17:       omp_if.else21:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_IF_END22]]
// CHECK17:       omp_if.end22:
// CHECK17-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
// CHECK17-NEXT:    ret i32 [[TMP108]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK17-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK17-SAME: (i32 signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
// CHECK17-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
// CHECK17-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
// CHECK17-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
// CHECK17-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
// CHECK17-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
// CHECK17-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
// CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
// CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
// CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
// CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
// CHECK17-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
// CHECK17-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
// CHECK17-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
// CHECK17-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
// CHECK17-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK17-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK17-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
// CHECK17-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK17-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK17:       omp_offload.failed.i:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR4]]
// CHECK17-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
// CHECK17:       .omp_outlined..1.exit:
// CHECK17-NEXT:    ret i32 0
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
// CHECK17-SAME: (i64 [[A:%.*]]) #[[ATTR3]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]]) #[[ATTR3]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK17-SAME: (i64 [[AA:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR3]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK17-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK17-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK17-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK17-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK17:       .cancel.exit:
// CHECK17-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK17-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK17:       .cancel.continue:
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK17-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
// CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR3]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK17-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK17-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK17-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
// CHECK17-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.omp_outlined..7 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]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR3]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK17-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK17-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
// CHECK17-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
// CHECK17-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
// CHECK17-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
// CHECK17-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK17-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
// CHECK17-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
// CHECK17-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
// CHECK17-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
// CHECK17-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
// CHECK17-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK17-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
// CHECK17-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
// CHECK17-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK17-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
// CHECK17-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK17-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
// CHECK17-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
// CHECK17-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
// CHECK17-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK17-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
// CHECK17-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
// CHECK17-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK17-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
// CHECK17-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
// CHECK17-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
// CHECK17-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
// CHECK17-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
// CHECK17-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z3bari
// CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CALL:%.*]] = call signext i32 @_Z3fooi(i32 signext [[TMP0]])
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP2]])
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
// CHECK17-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CALL3:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP4]])
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
// CHECK17-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CALL5:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP6]])
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
// CHECK17-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
// CHECK17-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
// CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
// CHECK17-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
// CHECK17-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
// CHECK17-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
// CHECK17-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK17:       omp_if.then:
// CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
// CHECK17-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
// CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
// CHECK17-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 8
// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
// CHECK17-NEXT:    store double* [[A]], double** [[TMP13]], align 8
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    store i64 8, i64* [[TMP14]], align 8
// CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP15]], align 8
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
// CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK17-NEXT:    store i64 4, i64* [[TMP20]], align 8
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK17-NEXT:    store i64 2, i64* [[TMP23]], align 8
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
// CHECK17-NEXT:    store i64 2, i64* [[TMP25]], align 8
// CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK17-NEXT:    store i64 8, i64* [[TMP26]], align 8
// CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP27]], align 8
// CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
// CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
// CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP31]], align 8
// CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK17-NEXT:    store i64 8, i64* [[TMP32]], align 8
// CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK17-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
// CHECK17-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
// CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK17-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
// CHECK17-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
// CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK17-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
// CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK17-NEXT:    store i8* null, i8** [[TMP39]], align 8
// CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK17-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK17-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK17:       omp_offload.failed:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK17:       omp_if.else:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_IF_END]]
// CHECK17:       omp_if.end:
// CHECK17-NEXT:    [[TMP45:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP45]]
// CHECK17-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK17-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK17-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP46]] to i32
// CHECK17-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
// CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP47]]
// CHECK17-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK17-NEXT:    ret i32 [[ADD4]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK17-NEXT:    [[AAA:%.*]] = alloca i8, align 1
// CHECK17-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK17-NEXT:    store i8 0, i8* [[AAA]], align 1
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK17-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
// CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK17:       omp_if.then:
// CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
// CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK17-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
// CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK17-NEXT:    store i8* null, i8** [[TMP26]], align 8
// CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK17-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK17-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK17:       omp_offload.failed:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK17:       omp_if.else:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_IF_END]]
// CHECK17:       omp_if.end:
// CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    ret i32 [[TMP31]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK17-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
// CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK17:       omp_if.then:
// CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP9]], align 8
// CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
// CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP14]], align 8
// CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
// CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
// CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
// CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK17-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK17:       omp_offload.failed:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK17:       omp_if.else:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK17-NEXT:    br label [[OMP_IF_END]]
// CHECK17:       omp_if.end:
// CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    ret i32 [[TMP24]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK17-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR3]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK17-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
// CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK17-NEXT:    store double [[ADD]], double* [[A]], align 8
// CHECK17-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
// CHECK17-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK17-NEXT:    store double [[INC]], double* [[A4]], align 8
// CHECK17-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
// CHECK17-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
// CHECK17-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK17-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK17-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK17-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
// CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
// CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK17-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK17-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK17-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
// CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
// CHECK17-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
// CHECK17-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK17-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK17-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK17-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK17-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK17-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK17-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK17-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK18-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
// CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
// CHECK18-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
// CHECK18-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
// CHECK18-NEXT:    [[A_CASTED3:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_CASTED5:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
// CHECK18-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [9 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [9 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [9 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
// CHECK18-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
// CHECK18-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
// CHECK18-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
// CHECK18-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
// CHECK18-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK18-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
// CHECK18-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
// CHECK18-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
// CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP7]])
// CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
// CHECK18-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i64 [[TMP12]]) #[[ATTR4:[0-9]+]]
// CHECK18-NEXT:    [[TMP13:%.*]] = load i16, i16* [[AA]], align 2
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP13]], i16* [[CONV2]], align 2
// CHECK18-NEXT:    [[TMP14:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
// CHECK18-NEXT:    store i64 [[TMP14]], i64* [[TMP16]], align 8
// CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK18-NEXT:    store i64 [[TMP14]], i64* [[TMP18]], align 8
// CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK18-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK18:       omp_offload.failed:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i64 [[TMP14]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[CONV4:%.*]] = bitcast i64* [[A_CASTED3]] to i32*
// CHECK18-NEXT:    store i32 [[TMP24]], i32* [[CONV4]], align 4
// CHECK18-NEXT:    [[TMP25:%.*]] = load i64, i64* [[A_CASTED3]], align 8
// CHECK18-NEXT:    [[TMP26:%.*]] = load i16, i16* [[AA]], align 2
// CHECK18-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED5]] to i16*
// CHECK18-NEXT:    store i16 [[TMP26]], i16* [[CONV6]], align 2
// CHECK18-NEXT:    [[TMP27:%.*]] = load i64, i64* [[AA_CASTED5]], align 8
// CHECK18-NEXT:    [[TMP28:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP28]], 10
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK18:       omp_if.then:
// CHECK18-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
// CHECK18-NEXT:    store i64 [[TMP25]], i64* [[TMP30]], align 8
// CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
// CHECK18-NEXT:    store i64 [[TMP25]], i64* [[TMP32]], align 8
// CHECK18-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK18-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
// CHECK18-NEXT:    store i64 [[TMP27]], i64* [[TMP35]], align 8
// CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK18-NEXT:    store i64 [[TMP27]], i64* [[TMP37]], align 8
// CHECK18-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP38]], align 8
// CHECK18-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK18-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
// CHECK18-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED10:%.*]], label [[OMP_OFFLOAD_CONT11:%.*]]
// CHECK18:       omp_offload.failed10:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT11]]
// CHECK18:       omp_offload.cont11:
// CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK18:       omp_if.else:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_IF_END]]
// CHECK18:       omp_if.end:
// CHECK18-NEXT:    [[TMP43:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
// CHECK18-NEXT:    store i32 [[TMP43]], i32* [[CONV13]], align 4
// CHECK18-NEXT:    [[TMP44:%.*]] = load i64, i64* [[A_CASTED12]], align 8
// CHECK18-NEXT:    [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CMP14:%.*]] = icmp sgt i32 [[TMP45]], 20
// CHECK18-NEXT:    br i1 [[CMP14]], label [[OMP_IF_THEN15:%.*]], label [[OMP_IF_ELSE21:%.*]]
// CHECK18:       omp_if.then15:
// CHECK18-NEXT:    [[TMP46:%.*]] = mul nuw i64 [[TMP2]], 4
// CHECK18-NEXT:    [[TMP47:%.*]] = mul nuw i64 5, [[TMP5]]
// CHECK18-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], 8
// CHECK18-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64*
// CHECK18-NEXT:    store i64 [[TMP44]], i64* [[TMP50]], align 8
// CHECK18-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
// CHECK18-NEXT:    store i64 [[TMP44]], i64* [[TMP52]], align 8
// CHECK18-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK18-NEXT:    store i64 4, i64* [[TMP53]], align 8
// CHECK18-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK18-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
// CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 8
// CHECK18-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
// CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 8
// CHECK18-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK18-NEXT:    store i64 40, i64* [[TMP59]], align 8
// CHECK18-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK18-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
// CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP62]], align 8
// CHECK18-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
// CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP64]], align 8
// CHECK18-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK18-NEXT:    store i64 8, i64* [[TMP65]], align 8
// CHECK18-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP66]], align 8
// CHECK18-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
// CHECK18-NEXT:    store float* [[VLA]], float** [[TMP68]], align 8
// CHECK18-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
// CHECK18-NEXT:    store float* [[VLA]], float** [[TMP70]], align 8
// CHECK18-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK18-NEXT:    store i64 [[TMP46]], i64* [[TMP71]], align 8
// CHECK18-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 3
// CHECK18-NEXT:    store i8* null, i8** [[TMP72]], align 8
// CHECK18-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 4
// CHECK18-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
// CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 8
// CHECK18-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 4
// CHECK18-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
// CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 8
// CHECK18-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK18-NEXT:    store i64 400, i64* [[TMP77]], align 8
// CHECK18-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 4
// CHECK18-NEXT:    store i8* null, i8** [[TMP78]], align 8
// CHECK18-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 5
// CHECK18-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
// CHECK18-NEXT:    store i64 5, i64* [[TMP80]], align 8
// CHECK18-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 5
// CHECK18-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
// CHECK18-NEXT:    store i64 5, i64* [[TMP82]], align 8
// CHECK18-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK18-NEXT:    store i64 8, i64* [[TMP83]], align 8
// CHECK18-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 5
// CHECK18-NEXT:    store i8* null, i8** [[TMP84]], align 8
// CHECK18-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 6
// CHECK18-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
// CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP86]], align 8
// CHECK18-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 6
// CHECK18-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
// CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP88]], align 8
// CHECK18-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK18-NEXT:    store i64 8, i64* [[TMP89]], align 8
// CHECK18-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 6
// CHECK18-NEXT:    store i8* null, i8** [[TMP90]], align 8
// CHECK18-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 7
// CHECK18-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
// CHECK18-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 8
// CHECK18-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 7
// CHECK18-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
// CHECK18-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 8
// CHECK18-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK18-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 8
// CHECK18-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 7
// CHECK18-NEXT:    store i8* null, i8** [[TMP96]], align 8
// CHECK18-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 8
// CHECK18-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
// CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 8
// CHECK18-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 8
// CHECK18-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
// CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 8
// CHECK18-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
// CHECK18-NEXT:    store i64 16, i64* [[TMP101]], align 8
// CHECK18-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 8
// CHECK18-NEXT:    store i8* null, i8** [[TMP102]], align 8
// CHECK18-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK18-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
// CHECK18-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
// CHECK18:       omp_offload.failed19:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
// CHECK18:       omp_offload.cont20:
// CHECK18-NEXT:    br label [[OMP_IF_END22:%.*]]
// CHECK18:       omp_if.else21:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_IF_END22]]
// CHECK18:       omp_if.end22:
// CHECK18-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
// CHECK18-NEXT:    ret i32 [[TMP108]]
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK18-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK18-SAME: (i32 signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
// CHECK18-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
// CHECK18-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
// CHECK18-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
// CHECK18-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
// CHECK18-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
// CHECK18-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
// CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
// CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
// CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
// CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
// CHECK18-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
// CHECK18-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
// CHECK18-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
// CHECK18-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
// CHECK18-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK18-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
// CHECK18-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
// CHECK18-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK18-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK18:       omp_offload.failed.i:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR4]]
// CHECK18-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
// CHECK18:       .omp_outlined..1.exit:
// CHECK18-NEXT:    ret i32 0
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
// CHECK18-SAME: (i64 [[A:%.*]]) #[[ATTR3]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]]) #[[ATTR3]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK18-SAME: (i64 [[AA:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK18-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR3]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK18-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK18-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK18-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK18-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK18-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK18:       .cancel.exit:
// CHECK18-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK18-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK18:       .cancel.continue:
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK18-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
// CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR3]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK18-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK18-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK18-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK18-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK18-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK18-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK18-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK18-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK18-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK18-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
// CHECK18-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.omp_outlined..7 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]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR3]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK18-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK18-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK18-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK18-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK18-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK18-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK18-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK18-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK18-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
// CHECK18-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
// CHECK18-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
// CHECK18-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
// CHECK18-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK18-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
// CHECK18-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
// CHECK18-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
// CHECK18-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
// CHECK18-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
// CHECK18-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK18-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
// CHECK18-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
// CHECK18-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK18-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
// CHECK18-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK18-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
// CHECK18-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
// CHECK18-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
// CHECK18-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK18-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
// CHECK18-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
// CHECK18-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK18-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
// CHECK18-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
// CHECK18-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
// CHECK18-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
// CHECK18-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
// CHECK18-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@_Z3bari
// CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CALL:%.*]] = call signext i32 @_Z3fooi(i32 signext [[TMP0]])
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP2]])
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
// CHECK18-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CALL3:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP4]])
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
// CHECK18-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CALL5:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP6]])
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
// CHECK18-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
// CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK18-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
// CHECK18-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
// CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
// CHECK18-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
// CHECK18-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
// CHECK18-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
// CHECK18-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
// CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
// CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK18:       omp_if.then:
// CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
// CHECK18-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
// CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
// CHECK18-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 8
// CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
// CHECK18-NEXT:    store double* [[A]], double** [[TMP13]], align 8
// CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK18-NEXT:    store i64 8, i64* [[TMP14]], align 8
// CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP15]], align 8
// CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
// CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
// CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK18-NEXT:    store i64 4, i64* [[TMP20]], align 8
// CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
// CHECK18-NEXT:    store i64 2, i64* [[TMP23]], align 8
// CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
// CHECK18-NEXT:    store i64 2, i64* [[TMP25]], align 8
// CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK18-NEXT:    store i64 8, i64* [[TMP26]], align 8
// CHECK18-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP27]], align 8
// CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
// CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
// CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP31]], align 8
// CHECK18-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK18-NEXT:    store i64 8, i64* [[TMP32]], align 8
// CHECK18-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK18-NEXT:    store i8* null, i8** [[TMP33]], align 8
// CHECK18-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK18-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
// CHECK18-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
// CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK18-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
// CHECK18-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
// CHECK18-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK18-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
// CHECK18-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK18-NEXT:    store i8* null, i8** [[TMP39]], align 8
// CHECK18-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK18-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK18-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK18:       omp_offload.failed:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK18:       omp_if.else:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_IF_END]]
// CHECK18:       omp_if.end:
// CHECK18-NEXT:    [[TMP45:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP45]]
// CHECK18-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK18-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK18-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP46]] to i32
// CHECK18-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
// CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP47]]
// CHECK18-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK18-NEXT:    ret i32 [[ADD4]]
//
//
// CHECK18-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK18-NEXT:    [[AAA:%.*]] = alloca i8, align 1
// CHECK18-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK18-NEXT:    store i8 0, i8* [[AAA]], align 1
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK18-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
// CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK18:       omp_if.then:
// CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
// CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
// CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP11]], align 8
// CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
// CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
// CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP16]], align 8
// CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
// CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
// CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
// CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP21]], align 8
// CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
// CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK18-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
// CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK18-NEXT:    store i8* null, i8** [[TMP26]], align 8
// CHECK18-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK18-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK18-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK18:       omp_offload.failed:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK18:       omp_if.else:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_IF_END]]
// CHECK18:       omp_if.end:
// CHECK18-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    ret i32 [[TMP31]]
//
//
// CHECK18-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK18-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
// CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK18:       omp_if.then:
// CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
// CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
// CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
// CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
// CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP9]], align 8
// CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
// CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
// CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
// CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
// CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP14]], align 8
// CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
// CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
// CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
// CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP19]], align 8
// CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK18-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK18:       omp_offload.failed:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK18:       omp_if.else:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK18-NEXT:    br label [[OMP_IF_END]]
// CHECK18:       omp_if.end:
// CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    ret i32 [[TMP24]]
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK18-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK18-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK18-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR3]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK18-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK18-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK18-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
// CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK18-NEXT:    store double [[ADD]], double* [[A]], align 8
// CHECK18-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
// CHECK18-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK18-NEXT:    store double [[INC]], double* [[A4]], align 8
// CHECK18-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
// CHECK18-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
// CHECK18-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK18-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK18-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
// CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK18-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK18-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
// CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
// CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK18-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK18-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK18-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
// CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
// CHECK18-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
// CHECK18-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK18-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK18-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK18-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK18-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK18-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK18-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK18-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK18-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK19-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
// CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
// CHECK19-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
// CHECK19-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_CASTED3:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
// CHECK19-NEXT:    [[A_CASTED10:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [9 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [9 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [9 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK19-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK19-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK19-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK19-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
// CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP5]])
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i32 [[TMP10]]) #[[ATTR4:[0-9]+]]
// CHECK19-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK19-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
// CHECK19-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
// CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP18]], i8** [[TMP19]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK19-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK19-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK19:       omp_offload.failed:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i32 [[TMP12]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    store i32 [[TMP22]], i32* [[A_CASTED2]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A_CASTED2]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = load i16, i16* [[AA]], align 2
// CHECK19-NEXT:    [[CONV4:%.*]] = bitcast i32* [[AA_CASTED3]] to i16*
// CHECK19-NEXT:    store i16 [[TMP24]], i16* [[CONV4]], align 2
// CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[AA_CASTED3]], align 4
// CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP26]], 10
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK19:       omp_if.then:
// CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
// CHECK19-NEXT:    store i32 [[TMP23]], i32* [[TMP28]], align 4
// CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK19-NEXT:    store i32 [[TMP23]], i32* [[TMP30]], align 4
// CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32*
// CHECK19-NEXT:    store i32 [[TMP25]], i32* [[TMP33]], align 4
// CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
// CHECK19-NEXT:    store i32 [[TMP25]], i32* [[TMP35]], align 4
// CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP36]], align 4
// CHECK19-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP37]], i8** [[TMP38]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK19-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK19-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK19:       omp_offload.failed8:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK19:       omp_offload.cont9:
// CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK19:       omp_if.else:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_IF_END]]
// CHECK19:       omp_if.end:
// CHECK19-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    store i32 [[TMP41]], i32* [[A_CASTED10]], align 4
// CHECK19-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A_CASTED10]], align 4
// CHECK19-NEXT:    [[TMP43:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], 20
// CHECK19-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN12:%.*]], label [[OMP_IF_ELSE18:%.*]]
// CHECK19:       omp_if.then12:
// CHECK19-NEXT:    [[TMP44:%.*]] = mul nuw i32 [[TMP1]], 4
// CHECK19-NEXT:    [[TMP45:%.*]] = sext i32 [[TMP44]] to i64
// CHECK19-NEXT:    [[TMP46:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK19-NEXT:    [[TMP47:%.*]] = mul nuw i32 [[TMP46]], 8
// CHECK19-NEXT:    [[TMP48:%.*]] = sext i32 [[TMP47]] to i64
// CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
// CHECK19-NEXT:    store i32 [[TMP42]], i32* [[TMP50]], align 4
// CHECK19-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK19-NEXT:    store i32 [[TMP42]], i32* [[TMP52]], align 4
// CHECK19-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    store i64 4, i64* [[TMP53]], align 4
// CHECK19-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK19-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
// CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 4
// CHECK19-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
// CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 4
// CHECK19-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK19-NEXT:    store i64 40, i64* [[TMP59]], align 4
// CHECK19-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP60]], align 4
// CHECK19-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP62]], align 4
// CHECK19-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32*
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP64]], align 4
// CHECK19-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK19-NEXT:    store i64 4, i64* [[TMP65]], align 4
// CHECK19-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP66]], align 4
// CHECK19-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
// CHECK19-NEXT:    store float* [[VLA]], float** [[TMP68]], align 4
// CHECK19-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
// CHECK19-NEXT:    store float* [[VLA]], float** [[TMP70]], align 4
// CHECK19-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK19-NEXT:    store i64 [[TMP45]], i64* [[TMP71]], align 4
// CHECK19-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 3
// CHECK19-NEXT:    store i8* null, i8** [[TMP72]], align 4
// CHECK19-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 4
// CHECK19-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
// CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 4
// CHECK19-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 4
// CHECK19-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
// CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 4
// CHECK19-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK19-NEXT:    store i64 400, i64* [[TMP77]], align 4
// CHECK19-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 4
// CHECK19-NEXT:    store i8* null, i8** [[TMP78]], align 4
// CHECK19-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 5
// CHECK19-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
// CHECK19-NEXT:    store i32 5, i32* [[TMP80]], align 4
// CHECK19-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 5
// CHECK19-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
// CHECK19-NEXT:    store i32 5, i32* [[TMP82]], align 4
// CHECK19-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK19-NEXT:    store i64 4, i64* [[TMP83]], align 4
// CHECK19-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 5
// CHECK19-NEXT:    store i8* null, i8** [[TMP84]], align 4
// CHECK19-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 6
// CHECK19-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP86]], align 4
// CHECK19-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 6
// CHECK19-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP88]], align 4
// CHECK19-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK19-NEXT:    store i64 4, i64* [[TMP89]], align 4
// CHECK19-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 6
// CHECK19-NEXT:    store i8* null, i8** [[TMP90]], align 4
// CHECK19-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 7
// CHECK19-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
// CHECK19-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 4
// CHECK19-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 7
// CHECK19-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
// CHECK19-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 4
// CHECK19-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK19-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 4
// CHECK19-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 7
// CHECK19-NEXT:    store i8* null, i8** [[TMP96]], align 4
// CHECK19-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 8
// CHECK19-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
// CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 4
// CHECK19-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 8
// CHECK19-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
// CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 4
// CHECK19-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
// CHECK19-NEXT:    store i64 12, i64* [[TMP101]], align 4
// CHECK19-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 8
// CHECK19-NEXT:    store i8* null, i8** [[TMP102]], align 4
// CHECK19-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK19-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
// CHECK19-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK19:       omp_offload.failed16:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK19:       omp_offload.cont17:
// CHECK19-NEXT:    br label [[OMP_IF_END19:%.*]]
// CHECK19:       omp_if.else18:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_IF_END19]]
// CHECK19:       omp_if.end19:
// CHECK19-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
// CHECK19-NEXT:    ret i32 [[TMP108]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK19-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK19-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
// CHECK19-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
// CHECK19-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
// CHECK19-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
// CHECK19-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
// CHECK19-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
// CHECK19-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
// CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
// CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
// CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
// CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
// CHECK19-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
// CHECK19-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
// CHECK19-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
// CHECK19-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
// CHECK19-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK19-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK19-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
// CHECK19-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK19-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK19:       omp_offload.failed.i:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR4]]
// CHECK19-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
// CHECK19:       .omp_outlined..1.exit:
// CHECK19-NEXT:    ret i32 0
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
// CHECK19-SAME: (i32 [[A:%.*]]) #[[ATTR3]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]]) #[[ATTR3]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK19-SAME: (i32 [[AA:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK19-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR3]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK19-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK19-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK19-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK19-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK19-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK19:       .cancel.exit:
// CHECK19-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK19-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK19:       .cancel.continue:
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK19-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR3]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK19-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK19-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK19-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.omp_outlined..7 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]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR3]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
// CHECK19-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK19-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
// CHECK19-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
// CHECK19-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK19-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
// CHECK19-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
// CHECK19-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
// CHECK19-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
// CHECK19-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
// CHECK19-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK19-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
// CHECK19-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK19-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
// CHECK19-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK19-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
// CHECK19-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
// CHECK19-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
// CHECK19-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK19-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
// CHECK19-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
// CHECK19-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK19-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
// CHECK19-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
// CHECK19-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
// CHECK19-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
// CHECK19-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
// CHECK19-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@_Z3bari
// CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooi(i32 [[TMP0]])
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CALL1:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
// CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CALL3:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP4]])
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
// CHECK19-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CALL5:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP6]])
// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
// CHECK19-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
// CHECK19-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
// CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK19-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
// CHECK19-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
// CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK19:       omp_if.then:
// CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
// CHECK19-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
// CHECK19-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
// CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
// CHECK19-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
// CHECK19-NEXT:    store double* [[A]], double** [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    store i64 8, i64* [[TMP14]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP15]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK19-NEXT:    store i64 4, i64* [[TMP20]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK19-NEXT:    store i32 2, i32* [[TMP23]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
// CHECK19-NEXT:    store i32 2, i32* [[TMP25]], align 4
// CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK19-NEXT:    store i64 4, i64* [[TMP26]], align 4
// CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP27]], align 4
// CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
// CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP31]], align 4
// CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK19-NEXT:    store i64 4, i64* [[TMP32]], align 4
// CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK19-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK19-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
// CHECK19-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
// CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK19-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
// CHECK19-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
// CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK19-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 4
// CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK19-NEXT:    store i8* null, i8** [[TMP39]], align 4
// CHECK19-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK19-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK19-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK19:       omp_offload.failed:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK19:       omp_if.else:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_IF_END]]
// CHECK19:       omp_if.end:
// CHECK19-NEXT:    [[TMP45:%.*]] = mul nsw i32 1, [[TMP1]]
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP45]]
// CHECK19-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK19-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP46]] to i32
// CHECK19-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
// CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP47]]
// CHECK19-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK19-NEXT:    ret i32 [[ADD3]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK19-NEXT:    [[AAA:%.*]] = alloca i8, align 1
// CHECK19-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK19-NEXT:    store i8 0, i8* [[AAA]], align 1
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
// CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK19-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK19:       omp_if.then:
// CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK19-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
// CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK19-NEXT:    store i8* null, i8** [[TMP26]], align 4
// CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK19-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK19-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK19:       omp_offload.failed:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK19:       omp_if.else:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_IF_END]]
// CHECK19:       omp_if.end:
// CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    ret i32 [[TMP31]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK19-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
// CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK19:       omp_if.then:
// CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP9]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
// CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP14]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
// CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
// CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP19]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK19-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK19:       omp_offload.failed:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK19:       omp_if.else:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK19-NEXT:    br label [[OMP_IF_END]]
// CHECK19:       omp_if.end:
// CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    ret i32 [[TMP24]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK19-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR3]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK19-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK19-NEXT:    store double [[ADD]], double* [[A]], align 4
// CHECK19-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
// CHECK19-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK19-NEXT:    store double [[INC]], double* [[A3]], align 4
// CHECK19-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
// CHECK19-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
// CHECK19-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK19-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK19-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK19-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK19-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK19-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK19-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK19-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
// CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
// CHECK19-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
// CHECK19-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK19-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK19-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK19-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK19-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK19-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK19-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK19-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@_Z3fooi
// CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK20-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
// CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
// CHECK20-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
// CHECK20-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_CASTED3:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
// CHECK20-NEXT:    [[A_CASTED10:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [9 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [9 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [9 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK20-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK20-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK20-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
// CHECK20-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
// CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP5]])
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i32 [[TMP10]]) #[[ATTR4:[0-9]+]]
// CHECK20-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
// CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK20-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
// CHECK20-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
// CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP17]], align 4
// CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP18]], i8** [[TMP19]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK20-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
// CHECK20-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK20:       omp_offload.failed:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i32 [[TMP12]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    store i32 [[TMP22]], i32* [[A_CASTED2]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A_CASTED2]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = load i16, i16* [[AA]], align 2
// CHECK20-NEXT:    [[CONV4:%.*]] = bitcast i32* [[AA_CASTED3]] to i16*
// CHECK20-NEXT:    store i16 [[TMP24]], i16* [[CONV4]], align 2
// CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[AA_CASTED3]], align 4
// CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP26]], 10
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK20:       omp_if.then:
// CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
// CHECK20-NEXT:    store i32 [[TMP23]], i32* [[TMP28]], align 4
// CHECK20-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK20-NEXT:    store i32 [[TMP23]], i32* [[TMP30]], align 4
// CHECK20-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP31]], align 4
// CHECK20-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32*
// CHECK20-NEXT:    store i32 [[TMP25]], i32* [[TMP33]], align 4
// CHECK20-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
// CHECK20-NEXT:    store i32 [[TMP25]], i32* [[TMP35]], align 4
// CHECK20-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP36]], align 4
// CHECK20-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP37]], i8** [[TMP38]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK20-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
// CHECK20-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
// CHECK20:       omp_offload.failed8:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
// CHECK20:       omp_offload.cont9:
// CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK20:       omp_if.else:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_IF_END]]
// CHECK20:       omp_if.end:
// CHECK20-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    store i32 [[TMP41]], i32* [[A_CASTED10]], align 4
// CHECK20-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A_CASTED10]], align 4
// CHECK20-NEXT:    [[TMP43:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], 20
// CHECK20-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN12:%.*]], label [[OMP_IF_ELSE18:%.*]]
// CHECK20:       omp_if.then12:
// CHECK20-NEXT:    [[TMP44:%.*]] = mul nuw i32 [[TMP1]], 4
// CHECK20-NEXT:    [[TMP45:%.*]] = sext i32 [[TMP44]] to i64
// CHECK20-NEXT:    [[TMP46:%.*]] = mul nuw i32 5, [[TMP3]]
// CHECK20-NEXT:    [[TMP47:%.*]] = mul nuw i32 [[TMP46]], 8
// CHECK20-NEXT:    [[TMP48:%.*]] = sext i32 [[TMP47]] to i64
// CHECK20-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
// CHECK20-NEXT:    store i32 [[TMP42]], i32* [[TMP50]], align 4
// CHECK20-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK20-NEXT:    store i32 [[TMP42]], i32* [[TMP52]], align 4
// CHECK20-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK20-NEXT:    store i64 4, i64* [[TMP53]], align 4
// CHECK20-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK20-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
// CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 4
// CHECK20-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
// CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 4
// CHECK20-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK20-NEXT:    store i64 40, i64* [[TMP59]], align 4
// CHECK20-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP60]], align 4
// CHECK20-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP62]], align 4
// CHECK20-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32*
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP64]], align 4
// CHECK20-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK20-NEXT:    store i64 4, i64* [[TMP65]], align 4
// CHECK20-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP66]], align 4
// CHECK20-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
// CHECK20-NEXT:    store float* [[VLA]], float** [[TMP68]], align 4
// CHECK20-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
// CHECK20-NEXT:    store float* [[VLA]], float** [[TMP70]], align 4
// CHECK20-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK20-NEXT:    store i64 [[TMP45]], i64* [[TMP71]], align 4
// CHECK20-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 3
// CHECK20-NEXT:    store i8* null, i8** [[TMP72]], align 4
// CHECK20-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 4
// CHECK20-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
// CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 4
// CHECK20-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 4
// CHECK20-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
// CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 4
// CHECK20-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK20-NEXT:    store i64 400, i64* [[TMP77]], align 4
// CHECK20-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 4
// CHECK20-NEXT:    store i8* null, i8** [[TMP78]], align 4
// CHECK20-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 5
// CHECK20-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
// CHECK20-NEXT:    store i32 5, i32* [[TMP80]], align 4
// CHECK20-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 5
// CHECK20-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
// CHECK20-NEXT:    store i32 5, i32* [[TMP82]], align 4
// CHECK20-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
// CHECK20-NEXT:    store i64 4, i64* [[TMP83]], align 4
// CHECK20-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 5
// CHECK20-NEXT:    store i8* null, i8** [[TMP84]], align 4
// CHECK20-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 6
// CHECK20-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP86]], align 4
// CHECK20-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 6
// CHECK20-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP88]], align 4
// CHECK20-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
// CHECK20-NEXT:    store i64 4, i64* [[TMP89]], align 4
// CHECK20-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 6
// CHECK20-NEXT:    store i8* null, i8** [[TMP90]], align 4
// CHECK20-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 7
// CHECK20-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
// CHECK20-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 4
// CHECK20-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 7
// CHECK20-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
// CHECK20-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 4
// CHECK20-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
// CHECK20-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 4
// CHECK20-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 7
// CHECK20-NEXT:    store i8* null, i8** [[TMP96]], align 4
// CHECK20-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 8
// CHECK20-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
// CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 4
// CHECK20-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 8
// CHECK20-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
// CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 4
// CHECK20-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
// CHECK20-NEXT:    store i64 12, i64* [[TMP101]], align 4
// CHECK20-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 8
// CHECK20-NEXT:    store i8* null, i8** [[TMP102]], align 4
// CHECK20-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK20-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
// CHECK20-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
// CHECK20:       omp_offload.failed16:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
// CHECK20:       omp_offload.cont17:
// CHECK20-NEXT:    br label [[OMP_IF_END19:%.*]]
// CHECK20:       omp_if.else18:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_IF_END19]]
// CHECK20:       omp_if.end19:
// CHECK20-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
// CHECK20-NEXT:    ret i32 [[TMP108]]
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK20-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_task_entry.
// CHECK20-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
// CHECK20-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
// CHECK20-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
// CHECK20-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
// CHECK20-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
// CHECK20-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
// CHECK20-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
// CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
// CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
// CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
// CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
// CHECK20-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
// CHECK20-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
// CHECK20-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
// CHECK20-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
// CHECK20-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK20-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
// CHECK20-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
// CHECK20-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK20-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
// CHECK20:       omp_offload.failed.i:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR4]]
// CHECK20-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
// CHECK20:       .omp_outlined..1.exit:
// CHECK20-NEXT:    ret i32 0
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
// CHECK20-SAME: (i32 [[A:%.*]]) #[[ATTR3]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]]) #[[ATTR3]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK20-SAME: (i32 [[AA:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK20-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR3]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK20-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK20-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK20-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK20-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK20-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK20:       .cancel.exit:
// CHECK20-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK20-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK20:       .cancel.continue:
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK20-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR3]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK20-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK20-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK20-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK20-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK20-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK20-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.omp_outlined..7 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]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR3]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK20-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK20-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK20-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
// CHECK20-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK20-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
// CHECK20-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
// CHECK20-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK20-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
// CHECK20-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
// CHECK20-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
// CHECK20-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
// CHECK20-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
// CHECK20-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK20-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
// CHECK20-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK20-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
// CHECK20-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK20-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
// CHECK20-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
// CHECK20-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
// CHECK20-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK20-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
// CHECK20-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
// CHECK20-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK20-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
// CHECK20-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
// CHECK20-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
// CHECK20-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
// CHECK20-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
// CHECK20-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@_Z3bari
// CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooi(i32 [[TMP0]])
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CALL1:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP2]])
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
// CHECK20-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CALL3:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP4]])
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
// CHECK20-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CALL5:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP6]])
// CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
// CHECK20-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK20-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
// CHECK20-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
// CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
// CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
// CHECK20-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
// CHECK20-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
// CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK20:       omp_if.then:
// CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
// CHECK20-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
// CHECK20-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
// CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
// CHECK20-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
// CHECK20-NEXT:    store double* [[A]], double** [[TMP13]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK20-NEXT:    store i64 8, i64* [[TMP14]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP15]], align 4
// CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
// CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
// CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
// CHECK20-NEXT:    store i64 4, i64* [[TMP20]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
// CHECK20-NEXT:    store i32 2, i32* [[TMP23]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
// CHECK20-NEXT:    store i32 2, i32* [[TMP25]], align 4
// CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
// CHECK20-NEXT:    store i64 4, i64* [[TMP26]], align 4
// CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP27]], align 4
// CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
// CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP31]], align 4
// CHECK20-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
// CHECK20-NEXT:    store i64 4, i64* [[TMP32]], align 4
// CHECK20-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK20-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK20-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK20-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
// CHECK20-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
// CHECK20-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK20-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
// CHECK20-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
// CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
// CHECK20-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 4
// CHECK20-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK20-NEXT:    store i8* null, i8** [[TMP39]], align 4
// CHECK20-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK20-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
// CHECK20-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK20:       omp_offload.failed:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK20:       omp_if.else:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_IF_END]]
// CHECK20:       omp_if.end:
// CHECK20-NEXT:    [[TMP45:%.*]] = mul nsw i32 1, [[TMP1]]
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP45]]
// CHECK20-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK20-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
// CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP46]] to i32
// CHECK20-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
// CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP47]]
// CHECK20-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
// CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK20-NEXT:    ret i32 [[ADD3]]
//
//
// CHECK20-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK20-NEXT:    [[AAA:%.*]] = alloca i8, align 1
// CHECK20-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK20-NEXT:    store i8 0, i8* [[AAA]], align 1
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
// CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK20-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK20:       omp_if.then:
// CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP11]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
// CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP16]], align 4
// CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
// CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP21]], align 4
// CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK20-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
// CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK20-NEXT:    store i8* null, i8** [[TMP26]], align 4
// CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK20-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK20-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK20:       omp_offload.failed:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK20:       omp_if.else:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_IF_END]]
// CHECK20:       omp_if.end:
// CHECK20-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    ret i32 [[TMP31]]
//
//
// CHECK20-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
// CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
// CHECK20-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
// CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
// CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK20:       omp_if.then:
// CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP9]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
// CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP14]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
// CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
// CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
// CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP19]], align 4
// CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
// CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
// CHECK20-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK20:       omp_offload.failed:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK20:       omp_if.else:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
// CHECK20-NEXT:    br label [[OMP_IF_END]]
// CHECK20:       omp_if.end:
// CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    ret i32 [[TMP24]]
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK20-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK20-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR3]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK20-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK20-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK20-NEXT:    store double [[ADD]], double* [[A]], align 4
// CHECK20-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
// CHECK20-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK20-NEXT:    store double [[INC]], double* [[A3]], align 4
// CHECK20-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
// CHECK20-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
// CHECK20-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK20-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK20-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK20-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK20-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK20-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK20-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK20-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
// CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
// CHECK20-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
// CHECK20-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK20-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK20-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..14
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR3]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
// CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK20-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK20-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK20-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK20-SAME: () #[[ATTR7:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK20-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK25-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK25-SAME: (i64 [[AA:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK25-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK25-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR1]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK25-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK25-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK25-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK25-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK25-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK25-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK25:       .cancel.exit:
// CHECK25-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK25-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK25:       .cancel.continue:
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK25-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK25-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK25-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
// CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK25-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
// CHECK25-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR1]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK25-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK25-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK25-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK25-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK25-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK25-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK25-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK25-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK25-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK25-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK25-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK25-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK25-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK25-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
// CHECK25-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR1]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK25-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK25-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK25-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK25-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK25-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK25-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK25-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK25-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK25-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK25-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK25-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
// CHECK25-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
// CHECK25-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
// CHECK25-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
// CHECK25-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK25-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
// CHECK25-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
// CHECK25-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
// CHECK25-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
// CHECK25-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
// CHECK25-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK25-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
// CHECK25-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
// CHECK25-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK25-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
// CHECK25-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK25-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
// CHECK25-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
// CHECK25-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
// CHECK25-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK25-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
// CHECK25-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK25-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
// CHECK25-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK25-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
// CHECK25-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK25-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
// CHECK25-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
// CHECK25-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
// CHECK25-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
// CHECK25-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK25-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK25-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
// CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK25-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
// CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK25-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK25-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
// CHECK25-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
// CHECK25-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK25-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK25-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK25-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
// CHECK25-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
// CHECK25-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
// CHECK25-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
// CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK25-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK25-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK25-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK25-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK25-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK25-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK25-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
// CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK25-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK25-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK25-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
// CHECK25-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK25-NEXT:    store double [[ADD]], double* [[A]], align 8
// CHECK25-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK25-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
// CHECK25-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK25-NEXT:    store double [[INC]], double* [[A4]], align 8
// CHECK25-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
// CHECK25-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
// CHECK25-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK25-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK25-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK25-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
// CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK25-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK25-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK25-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK25-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK25-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
// CHECK25-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK26-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK26-SAME: (i64 [[AA:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK26-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK26-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[AA:%.*]]) #[[ATTR1]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK26-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK26-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK26-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK26-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK26-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK26-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK26:       .cancel.exit:
// CHECK26-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK26-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK26:       .cancel.continue:
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK26-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK26-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK26-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
// CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK26-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
// CHECK26-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR1]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK26-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK26-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK26-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK26-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK26-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK26-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK26-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK26-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK26-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK26-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK26-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK26-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK26-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK26-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK26-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
// CHECK26-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR1]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
// CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
// CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
// CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
// CHECK26-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
// CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK26-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
// CHECK26-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK26-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
// CHECK26-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
// CHECK26-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
// CHECK26-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
// CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
// CHECK26-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
// CHECK26-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
// CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
// CHECK26-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK26-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
// CHECK26-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
// CHECK26-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
// CHECK26-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
// CHECK26-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
// CHECK26-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
// CHECK26-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
// CHECK26-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
// CHECK26-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
// CHECK26-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
// CHECK26-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
// CHECK26-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
// CHECK26-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
// CHECK26-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK26-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
// CHECK26-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
// CHECK26-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
// CHECK26-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
// CHECK26-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
// CHECK26-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK26-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
// CHECK26-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK26-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
// CHECK26-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK26-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
// CHECK26-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK26-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
// CHECK26-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
// CHECK26-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
// CHECK26-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
// CHECK26-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK26-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK26-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
// CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK26-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
// CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK26-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
// CHECK26-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
// CHECK26-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
// CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
// CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
// CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
// CHECK26-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
// CHECK26-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
// CHECK26-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 8
// CHECK26-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
// CHECK26-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
// CHECK26-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
// CHECK26-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 8
// CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK26-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK26-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK26-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK26-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK26-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK26-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK26-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
// CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK26-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK26-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK26-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
// CHECK26-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK26-NEXT:    store double [[ADD]], double* [[A]], align 8
// CHECK26-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK26-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
// CHECK26-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK26-NEXT:    store double [[INC]], double* [[A4]], align 8
// CHECK26-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
// CHECK26-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
// CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
// CHECK26-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
// CHECK26-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK26-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK26-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
// CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
// CHECK26-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
// CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
// CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
// CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
// CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
// CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK26-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK26-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK26-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
// CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK26-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK26-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
// CHECK26-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK27-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK27-SAME: (i32 [[AA:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK27-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK27-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR1]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK27-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK27-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK27-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK27-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK27-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK27-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK27:       .cancel.exit:
// CHECK27-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK27-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK27:       .cancel.continue:
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK27-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK27-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR1]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK27-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK27-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK27-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK27-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK27-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK27-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK27-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK27-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK27-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK27-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK27-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK27-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR1]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK27-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK27-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK27-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK27-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK27-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK27-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK27-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK27-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK27-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK27-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
// CHECK27-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK27-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
// CHECK27-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
// CHECK27-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK27-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
// CHECK27-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
// CHECK27-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
// CHECK27-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
// CHECK27-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
// CHECK27-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK27-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
// CHECK27-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
// CHECK27-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK27-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
// CHECK27-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK27-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
// CHECK27-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
// CHECK27-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
// CHECK27-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK27-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
// CHECK27-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK27-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
// CHECK27-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK27-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
// CHECK27-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK27-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
// CHECK27-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
// CHECK27-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
// CHECK27-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
// CHECK27-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK27-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK27-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK27-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK27-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK27-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
// CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK27-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK27-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK27-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 4
// CHECK27-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK27-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
// CHECK27-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
// CHECK27-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
// CHECK27-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
// CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK27-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK27-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK27-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK27-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK27-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK27-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK27-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK27-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK27-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK27-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK27-NEXT:    store double [[ADD]], double* [[A]], align 4
// CHECK27-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK27-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
// CHECK27-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK27-NEXT:    store double [[INC]], double* [[A3]], align 4
// CHECK27-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
// CHECK27-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
// CHECK27-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK27-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK27-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK27-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
// CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK27-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK27-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK27-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK27-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
// CHECK27-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
// CHECK28-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
// CHECK28-SAME: (i32 [[AA:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK28-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK28-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[AA:%.*]]) #[[ATTR1]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK28-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK28-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
// CHECK28-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
// CHECK28-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 1)
// CHECK28-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP3]], 0
// CHECK28-NEXT:    br i1 [[TMP4]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
// CHECK28:       .cancel.exit:
// CHECK28-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP2]])
// CHECK28-NEXT:    br label [[DOTCANCEL_CONTINUE]]
// CHECK28:       .cancel.continue:
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
// CHECK28-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK28-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
// CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR1]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
// CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK28-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK28-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK28-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
// CHECK28-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK28-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK28-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK28-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK28-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK28-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK28-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK28-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK28-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK28-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR1]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
// CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
// CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
// CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
// CHECK28-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
// CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK28-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
// CHECK28-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK28-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
// CHECK28-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
// CHECK28-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
// CHECK28-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
// CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
// CHECK28-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
// CHECK28-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
// CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
// CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
// CHECK28-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
// CHECK28-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
// CHECK28-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
// CHECK28-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
// CHECK28-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
// CHECK28-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
// CHECK28-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
// CHECK28-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
// CHECK28-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
// CHECK28-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
// CHECK28-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
// CHECK28-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
// CHECK28-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
// CHECK28-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
// CHECK28-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
// CHECK28-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
// CHECK28-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
// CHECK28-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
// CHECK28-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
// CHECK28-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
// CHECK28-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
// CHECK28-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
// CHECK28-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
// CHECK28-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
// CHECK28-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
// CHECK28-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
// CHECK28-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
// CHECK28-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
// CHECK28-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
// CHECK28-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
// CHECK28-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
// CHECK28-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK28-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
// CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK28-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK28-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
// CHECK28-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
// CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
// CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
// CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
// CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK28-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
// CHECK28-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
// CHECK28-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 4
// CHECK28-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 4
// CHECK28-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
// CHECK28-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
// CHECK28-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
// CHECK28-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 4
// CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK28-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK28-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
// CHECK28-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK28-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK28-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK28-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
// CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR1]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
// CHECK28-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
// CHECK28-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
// CHECK28-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
// CHECK28-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK28-NEXT:    store double [[ADD]], double* [[A]], align 4
// CHECK28-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK28-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
// CHECK28-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
// CHECK28-NEXT:    store double [[INC]], double* [[A3]], align 4
// CHECK28-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
// CHECK28-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
// CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
// CHECK28-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
// CHECK28-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
// CHECK28-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
// CHECK28-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
// CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
// CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], 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]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
// CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
// CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
// CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
// CHECK28-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
// CHECK28-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
// CHECK28-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 4
// CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
// CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK28-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
// CHECK28-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
// CHECK28-NEXT:    ret void
//