// 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 6 target regions


// Check target registration is registered as a Ctor.


template<typename tx>
tx ftemplate(int n) {
  tx a = 0;

  #pragma omp target parallel num_threads(tx(20))
  {
  }

  short b = 1;
  #pragma omp target parallel num_threads(b)
  {
    a += b;
  }

  return a;
}

static
int fstatic(int n) {

  #pragma omp target parallel num_threads(n)
  {
  }

  #pragma omp target parallel num_threads(32+n)
  {
  }

  return n+1;
}

struct S1 {
  double a;

  int r1(int n){
    int b = 1;

    #pragma omp target parallel num_threads(n-b)
    {
      this->a = (double)b + 1.5;
    }

    #pragma omp target parallel num_threads(1024)
    {
      this->a = 2.5;
    }

    return (int)a;
  }
};

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

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

  a += fstatic(n);

  a += ftemplate<int>(n);

  return a;
}


// Check that the offloading functions are emitted and that the parallel function
// is appropriately guarded.


#endif
// CHECK1-LABEL: define {{[^@]+}}@_Z3bari
// 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:    [[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 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], 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 @_ZL7fstatici(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 @_Z9ftemplateIiET_i(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* [[A]], align 4
// CHECK1-NEXT:    ret i32 [[TMP6]]
//
//
// 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:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], 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:    store i32 1, i32* [[B]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
// CHECK1-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
// CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
// CHECK1-NEXT:    store double* [[A]], double** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP10]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK1-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK1-NEXT:    br i1 [[TMP25]], 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_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR3:[0-9]+]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
// CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
// CHECK1-NEXT:    store double* [[A3]], double** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
// CHECK1-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
// CHECK1-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK1:       omp_offload.failed7:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK1:       omp_offload.cont8:
// CHECK1-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP35:%.*]] = load double, double* [[A9]], align 8
// CHECK1-NEXT:    [[CONV10:%.*]] = fptosi double [[TMP35]] to i32
// CHECK1-NEXT:    ret i32 [[CONV10]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP4]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
// CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP7]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
// CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT:    br i1 [[TMP12]], 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_l104(i64 [[TMP2]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
// CHECK1-NEXT:    store i32 [[TMP14]], i32* [[CONV3]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP17]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK1-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK1-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK1-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK1:       omp_offload.failed7:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP15]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK1:       omp_offload.cont8:
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP26]], 1
// CHECK1-NEXT:    ret i32 [[ADD9]]
//
//
// 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:    [[B:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__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:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
// CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
// CHECK1-NEXT:    br i1 [[TMP1]], 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_l88() #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    store i16 1, i16* [[B]], align 2
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
// CHECK1-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP5]], i16* [[CONV1]], align 2
// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP7]], i16* [[CONV2]], align 2
// CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK1-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK1-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK1-NEXT:    store i64 [[TMP8]], i64* [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK1-NEXT:    store i64 [[TMP8]], i64* [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK1-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
// CHECK1-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
// CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK1-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK1:       omp_offload.failed3:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK1:       omp_offload.cont4:
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
// CHECK1-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
// CHECK1-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]]) #[[ATTR2:[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK1-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[B]], i64* [[B_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 i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
// CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 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:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK1-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK1-NEXT:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK1-NEXT:    store double 2.500000e+00, double* [[A]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK1-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..4 to void (i32*, i32*, ...)*))
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
// CHECK1-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..7 to void (i32*, i32*, ...)*))
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK1-SAME: () #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// 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..10 to void (i32*, i32*, ...)*))
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
// CHECK1-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
// CHECK1-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[B:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[B_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 [[B]], i64* [[B_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z3bari
// 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:    [[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 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], 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 @_ZL7fstatici(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 @_Z9ftemplateIiET_i(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* [[A]], align 4
// CHECK2-NEXT:    ret i32 [[TMP6]]
//
//
// 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:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], 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:    store i32 1, i32* [[B]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
// CHECK2-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
// CHECK2-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
// CHECK2-NEXT:    store double* [[A]], double** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP10]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK2-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK2-NEXT:    br i1 [[TMP25]], 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_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR3:[0-9]+]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
// CHECK2-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
// CHECK2-NEXT:    store double* [[A3]], double** [[TMP29]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP30]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
// CHECK2-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
// CHECK2-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK2:       omp_offload.failed7:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR3]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK2:       omp_offload.cont8:
// CHECK2-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP35:%.*]] = load double, double* [[A9]], align 8
// CHECK2-NEXT:    [[CONV10:%.*]] = fptosi double [[TMP35]] to i32
// CHECK2-NEXT:    ret i32 [[CONV10]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
// CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP4]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
// CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP7]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
// CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK2-NEXT:    br i1 [[TMP12]], 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_l104(i64 [[TMP2]]) #[[ATTR3]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
// CHECK2-NEXT:    store i32 [[TMP14]], i32* [[CONV3]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP17]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP19]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK2-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK2-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK2:       omp_offload.failed7:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP15]]) #[[ATTR3]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK2:       omp_offload.cont8:
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP26]], 1
// CHECK2-NEXT:    ret i32 [[ADD9]]
//
//
// 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:    [[B:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__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:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
// CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
// CHECK2-NEXT:    br i1 [[TMP1]], 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_l88() #[[ATTR3]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    store i16 1, i16* [[B]], align 2
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
// CHECK2-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP5]], i16* [[CONV1]], align 2
// CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP7]], i16* [[CONV2]], align 2
// CHECK2-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK2-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK2-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK2-NEXT:    store i64 [[TMP8]], i64* [[TMP20]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK2-NEXT:    store i64 [[TMP8]], i64* [[TMP22]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK2-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
// CHECK2-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
// CHECK2-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK2-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK2:       omp_offload.failed3:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR3]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK2:       omp_offload.cont4:
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
// CHECK2-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
// CHECK2-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
// CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]]) #[[ATTR2:[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK2-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[B]], i64* [[B_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 i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
// CHECK2-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 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:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK2-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[ATTR2]] {
// 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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK2-NEXT:    store double 2.500000e+00, double* [[A]], align 8
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK2-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..4 to void (i32*, i32*, ...)*))
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
// CHECK2-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..7 to void (i32*, i32*, ...)*))
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK2-SAME: () #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// 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..10 to void (i32*, i32*, ...)*))
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
// CHECK2-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
// CHECK2-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[B:%.*]]) #[[ATTR2]] {
// 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 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 [[B]], i64* [[B_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK2-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK2-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z3bari
// 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:    [[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 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], 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 @_ZL7fstatici(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 @_Z9ftemplateIiET_i(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* [[A]], align 4
// CHECK3-NEXT:    ret i32 [[TMP6]]
//
//
// 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:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], 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:    store i32 1, i32* [[B]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
// CHECK3-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
// CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
// CHECK3-NEXT:    store double* [[A]], double** [[TMP9]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP10]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK3-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK3-NEXT:    br i1 [[TMP25]], 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_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR3:[0-9]+]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
// CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
// CHECK3-NEXT:    store double* [[A2]], double** [[TMP29]], align 4
// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP30]], align 4
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
// CHECK3-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
// CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK3:       omp_offload.failed6:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK3:       omp_offload.cont7:
// CHECK3-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP35:%.*]] = load double, double* [[A8]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = fptosi double [[TMP35]] to i32
// CHECK3-NEXT:    ret i32 [[CONV]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP4]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
// CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP6]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP7]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
// CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK3-NEXT:    br i1 [[TMP12]], 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_l104(i32 [[TMP2]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP19]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK3-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK3-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK3:       omp_offload.failed6:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP15]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK3:       omp_offload.cont7:
// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP26]], 1
// CHECK3-NEXT:    ret i32 [[ADD8]]
//
//
// 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:    [[B:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__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:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
// CHECK3-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
// CHECK3-NEXT:    br i1 [[TMP1]], 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_l88() #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    store i16 1, i16* [[B]], align 2
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
// CHECK3-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP5]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP7]], i16* [[CONV1]], align 2
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK3-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK3-NEXT:    store i32 [[TMP6]], i32* [[TMP15]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK3-NEXT:    store i32 [[TMP6]], i32* [[TMP17]], align 4
// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP20]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP22]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK3-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
// CHECK3-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
// CHECK3-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK3-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
// CHECK3:       omp_offload.failed2:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR3]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
// CHECK3:       omp_offload.cont3:
// CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
// CHECK3-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
// CHECK3-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]]) #[[ATTR2:[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK3-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] 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:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK3-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK3-NEXT:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK3-NEXT:    store double 2.500000e+00, double* [[A]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK3-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..4 to void (i32*, i32*, ...)*))
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
// CHECK3-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..7 to void (i32*, i32*, ...)*))
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK3-SAME: () #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// 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..10 to void (i32*, i32*, ...)*))
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
// CHECK3-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK3-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[B:%.*]]) #[[ATTR2]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[B_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 [[B]], i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z3bari
// 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:    [[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 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], 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 @_ZL7fstatici(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 @_Z9ftemplateIiET_i(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* [[A]], align 4
// CHECK4-NEXT:    ret i32 [[TMP6]]
//
//
// 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:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], 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:    store i32 1, i32* [[B]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
// CHECK4-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
// CHECK4-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
// CHECK4-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
// CHECK4-NEXT:    store double* [[A]], double** [[TMP9]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP10]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP15]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK4-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK4-NEXT:    br i1 [[TMP25]], 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_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR3:[0-9]+]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
// CHECK4-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 4
// CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
// CHECK4-NEXT:    store double* [[A2]], double** [[TMP29]], align 4
// CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP30]], align 4
// CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
// CHECK4-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
// CHECK4-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK4:       omp_offload.failed6:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR3]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK4:       omp_offload.cont7:
// CHECK4-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP35:%.*]] = load double, double* [[A8]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = fptosi double [[TMP35]] to i32
// CHECK4-NEXT:    ret i32 [[CONV]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
// CHECK4-NEXT:    store i32 [[TMP2]], i32* [[TMP4]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
// CHECK4-NEXT:    store i32 [[TMP2]], i32* [[TMP6]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP7]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
// CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK4-NEXT:    br i1 [[TMP12]], 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_l104(i32 [[TMP2]]) #[[ATTR3]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK4-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK4-NEXT:    store i32 [[TMP15]], i32* [[TMP19]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK4-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK4-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK4:       omp_offload.failed6:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP15]]) #[[ATTR3]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK4:       omp_offload.cont7:
// CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP26]], 1
// CHECK4-NEXT:    ret i32 [[ADD8]]
//
//
// 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:    [[B:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__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:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
// CHECK4-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
// CHECK4-NEXT:    br i1 [[TMP1]], 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_l88() #[[ATTR3]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    store i16 1, i16* [[B]], align 2
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
// CHECK4-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP5]], i16* [[CONV]], align 2
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP7]], i16* [[CONV1]], align 2
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK4-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
// CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK4-NEXT:    store i32 [[TMP6]], i32* [[TMP15]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK4-NEXT:    store i32 [[TMP6]], i32* [[TMP17]], align 4
// CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP22]], align 4
// CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK4-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
// CHECK4-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
// CHECK4-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK4-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
// CHECK4:       omp_offload.failed2:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR3]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
// CHECK4:       omp_offload.cont3:
// CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
// CHECK4-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
// CHECK4-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]]) #[[ATTR2:[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK4-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] 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:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK4-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[ATTR2]] {
// 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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK4-NEXT:    store double 2.500000e+00, double* [[A]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK4-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..4 to void (i32*, i32*, ...)*))
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
// CHECK4-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..7 to void (i32*, i32*, ...)*))
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK4-SAME: () #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// 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..10 to void (i32*, i32*, ...)*))
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
// CHECK4-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK4-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK4-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK4-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[B:%.*]]) #[[ATTR2]] {
// 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 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 [[B]], i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK4-SAME: () #[[ATTR4:[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]+}}__ZL7fstatici_l104
// CHECK9-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK9-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*, ...)*))
// 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]+}}__ZL7fstatici_l108
// CHECK9-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK9-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..1 to void (i32*, i32*, ...)*))
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__ZN2S12r1Ei_l121
// CHECK9-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK9-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[B]], i64* [[B_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 i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
// CHECK9-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 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:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK9-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK9-NEXT:    store double 2.500000e+00, double* [[A]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK9-SAME: () #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// CHECK9-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..4 to void (i32*, i32*, ...)*))
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__Z9ftemplateIiET_i_l93
// CHECK9-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK9-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK9-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK9-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
// CHECK9-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[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:    [[B_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 [[B]], i64* [[B_ADDR]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK9-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
// CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK10-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK10-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*, ...)*))
// 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]+}}__ZL7fstatici_l108
// CHECK10-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK10-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..1 to void (i32*, i32*, ...)*))
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__ZN2S12r1Ei_l121
// CHECK10-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
// CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK10-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[B]], i64* [[B_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 i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
// CHECK10-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 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:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK10-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK10-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK10-NEXT:    store double 2.500000e+00, double* [[A]], align 8
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK10-SAME: () #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// CHECK10-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..4 to void (i32*, i32*, ...)*))
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__Z9ftemplateIiET_i_l93
// CHECK10-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK10-NEXT:  entry:
// CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK10-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
// CHECK10-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK10-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
// CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK10-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK10-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
// CHECK10-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
// CHECK10-NEXT:    ret void
//
//
// CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[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:    [[B_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 [[B]], i64* [[B_ADDR]], align 8
// CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK10-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK10-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
// CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
// CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK10-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK11-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK11-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*, ...)*))
// 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]+}}__ZL7fstatici_l108
// CHECK11-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK11-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..1 to void (i32*, i32*, ...)*))
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__ZN2S12r1Ei_l121
// CHECK11-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK11-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] 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:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK11-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK11-NEXT:    store double 2.500000e+00, double* [[A]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK11-SAME: () #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// CHECK11-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..4 to void (i32*, i32*, ...)*))
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__Z9ftemplateIiET_i_l93
// CHECK11-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK11-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
// CHECK11-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK11-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK11-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK11-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[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:    [[B_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 [[B]], i32* [[B_ADDR]], align 4
// CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK11-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
// CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK12-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK12-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*, ...)*))
// 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]+}}__ZL7fstatici_l108
// CHECK12-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK12-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..1 to void (i32*, i32*, ...)*))
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__ZN2S12r1Ei_l121
// CHECK12-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK12-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] 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:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK12-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK12-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK12-NEXT:    store double 2.500000e+00, double* [[A]], align 4
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK12-SAME: () #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// CHECK12-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..4 to void (i32*, i32*, ...)*))
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__Z9ftemplateIiET_i_l93
// CHECK12-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK12-NEXT:  entry:
// CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK12-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
// CHECK12-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK12-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK12-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK12-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
// CHECK12-NEXT:    ret void
//
//
// CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[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:    [[B_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 [[B]], i32* [[B_ADDR]], align 4
// CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK12-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK12-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
// CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK12-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@_Z3bari
// 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:    [[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 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], 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 @_ZL7fstatici(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 @_Z9ftemplateIiET_i(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* [[A]], align 4
// CHECK17-NEXT:    ret i32 [[TMP6]]
//
//
// 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:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], 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:    store i32 1, i32* [[B]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
// CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
// CHECK17-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
// CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
// CHECK17-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
// CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
// CHECK17-NEXT:    store double* [[A]], double** [[TMP9]], align 8
// CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP10]], align 8
// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
// CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
// CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
// CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP15]], align 8
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
// CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK17-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK17-NEXT:    br i1 [[TMP25]], 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_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR3:[0-9]+]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
// CHECK17-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 8
// CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
// CHECK17-NEXT:    store double* [[A3]], double** [[TMP29]], align 8
// CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP30]], align 8
// CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
// CHECK17-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
// CHECK17-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK17:       omp_offload.failed7:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK17:       omp_offload.cont8:
// CHECK17-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP35:%.*]] = load double, double* [[A9]], align 8
// CHECK17-NEXT:    [[CONV10:%.*]] = fptosi double [[TMP35]] to i32
// CHECK17-NEXT:    ret i32 [[CONV10]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
// CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
// CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP4]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
// CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP6]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP7]], align 8
// CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK17-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
// CHECK17-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK17-NEXT:    br i1 [[TMP12]], 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_l104(i64 [[TMP2]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
// CHECK17-NEXT:    store i32 [[TMP14]], i32* [[CONV3]], align 4
// CHECK17-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK17-NEXT:    store i64 [[TMP15]], i64* [[TMP17]], align 8
// CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK17-NEXT:    store i64 [[TMP15]], i64* [[TMP19]], align 8
// CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK17-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK17-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK17-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK17:       omp_offload.failed7:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP15]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK17:       omp_offload.cont8:
// CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP26]], 1
// CHECK17-NEXT:    ret i32 [[ADD9]]
//
//
// 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:    [[B:%.*]] = alloca i16, align 2
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__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:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
// CHECK17-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
// CHECK17-NEXT:    br i1 [[TMP1]], 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_l88() #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK17:       omp_offload.cont:
// CHECK17-NEXT:    store i16 1, i16* [[B]], align 2
// CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
// CHECK17-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP5]], i16* [[CONV1]], align 2
// CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK17-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP7]], i16* [[CONV2]], align 2
// CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK17-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK17-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
// CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK17-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP15]], align 8
// CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK17-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK17-NEXT:    store i64 [[TMP8]], i64* [[TMP20]], align 8
// CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK17-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK17-NEXT:    store i64 [[TMP8]], i64* [[TMP22]], align 8
// CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK17-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK17-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK17-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
// CHECK17-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
// CHECK17-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK17-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK17:       omp_offload.failed3:
// CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR3]]
// CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK17:       omp_offload.cont4:
// CHECK17-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
// CHECK17-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
// CHECK17-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
// CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]]) #[[ATTR2:[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK17-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[B]], i64* [[B_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 i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
// CHECK17-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 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:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK17-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[ATTR2]] {
// 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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK17-NEXT:    store double 2.500000e+00, double* [[A]], align 8
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK17-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..4 to void (i32*, i32*, ...)*))
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
// CHECK17-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..7 to void (i32*, i32*, ...)*))
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK17-SAME: () #[[ATTR1]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// 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..10 to void (i32*, i32*, ...)*))
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
// CHECK17-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK17-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
// CHECK17-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK17-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
// CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK17-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK17-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
// CHECK17-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[B:%.*]]) #[[ATTR2]] {
// 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 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 [[B]], i64* [[B_ADDR]], align 8
// CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK17-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK17-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
// CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
// CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK17-NEXT:    ret void
//
//
// CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK17-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK17-NEXT:  entry:
// CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK17-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@_Z3bari
// 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:    [[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 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], 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 @_ZL7fstatici(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 @_Z9ftemplateIiET_i(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* [[A]], align 4
// CHECK18-NEXT:    ret i32 [[TMP6]]
//
//
// 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:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], 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:    store i32 1, i32* [[B]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
// CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
// CHECK18-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
// CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
// CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
// CHECK18-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
// CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
// CHECK18-NEXT:    store double* [[A]], double** [[TMP9]], align 8
// CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP10]], align 8
// CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
// CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
// CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
// CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP15]], align 8
// CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
// CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
// CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK18-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK18-NEXT:    br i1 [[TMP25]], 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_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR3:[0-9]+]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
// CHECK18-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 8
// CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
// CHECK18-NEXT:    store double* [[A3]], double** [[TMP29]], align 8
// CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP30]], align 8
// CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
// CHECK18-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
// CHECK18-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK18:       omp_offload.failed7:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK18:       omp_offload.cont8:
// CHECK18-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP35:%.*]] = load double, double* [[A9]], align 8
// CHECK18-NEXT:    [[CONV10:%.*]] = fptosi double [[TMP35]] to i32
// CHECK18-NEXT:    ret i32 [[CONV10]]
//
//
// CHECK18-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
// CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
// CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
// CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
// CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP4]], align 8
// CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
// CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP6]], align 8
// CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP7]], align 8
// CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK18-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
// CHECK18-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK18-NEXT:    br i1 [[TMP12]], 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_l104(i64 [[TMP2]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
// CHECK18-NEXT:    store i32 [[TMP14]], i32* [[CONV3]], align 4
// CHECK18-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
// CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
// CHECK18-NEXT:    store i64 [[TMP15]], i64* [[TMP17]], align 8
// CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
// CHECK18-NEXT:    store i64 [[TMP15]], i64* [[TMP19]], align 8
// CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP20]], align 8
// CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK18-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK18-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK18-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
// CHECK18:       omp_offload.failed7:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP15]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
// CHECK18:       omp_offload.cont8:
// CHECK18-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP26]], 1
// CHECK18-NEXT:    ret i32 [[ADD9]]
//
//
// 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:    [[B:%.*]] = alloca i16, align 2
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__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:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
// CHECK18-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
// CHECK18-NEXT:    br i1 [[TMP1]], 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_l88() #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK18:       omp_offload.cont:
// CHECK18-NEXT:    store i16 1, i16* [[B]], align 2
// CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
// CHECK18-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP5]], i16* [[CONV1]], align 2
// CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK18-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP7]], i16* [[CONV2]], align 2
// CHECK18-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
// CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
// CHECK18-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
// CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
// CHECK18-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
// CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK18-NEXT:    store i8* null, i8** [[TMP13]], align 8
// CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK18-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
// CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP15]], align 8
// CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], 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 [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK18-NEXT:    store i8* null, i8** [[TMP18]], align 8
// CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
// CHECK18-NEXT:    store i64 [[TMP8]], i64* [[TMP20]], align 8
// CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK18-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK18-NEXT:    store i64 [[TMP8]], i64* [[TMP22]], align 8
// CHECK18-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK18-NEXT:    store i8* null, i8** [[TMP23]], align 8
// CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK18-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK18-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
// CHECK18-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
// CHECK18-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK18-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK18:       omp_offload.failed3:
// CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR3]]
// CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK18:       omp_offload.cont4:
// CHECK18-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
// CHECK18-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
// CHECK18-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
// CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[B:%.*]]) #[[ATTR2:[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK18-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[B]], i64* [[B_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 i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
// CHECK18-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 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:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK18-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[ATTR2]] {
// 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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK18-NEXT:    store double 2.500000e+00, double* [[A]], align 8
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK18-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..4 to void (i32*, i32*, ...)*))
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
// CHECK18-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..7 to void (i32*, i32*, ...)*))
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK18-SAME: () #[[ATTR1]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// 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..10 to void (i32*, i32*, ...)*))
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
// CHECK18-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK18-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
// CHECK18-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK18-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
// CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK18-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK18-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
// CHECK18-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[B:%.*]]) #[[ATTR2]] {
// 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 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 [[B]], i64* [[B_ADDR]], align 8
// CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK18-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK18-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
// CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
// CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK18-NEXT:    ret void
//
//
// CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK18-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK18-NEXT:  entry:
// CHECK18-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK18-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@_Z3bari
// 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:    [[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 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], 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 @_ZL7fstatici(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 @_Z9ftemplateIiET_i(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* [[A]], align 4
// CHECK19-NEXT:    ret i32 [[TMP6]]
//
//
// 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:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], 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:    store i32 1, i32* [[B]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
// CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
// CHECK19-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
// CHECK19-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
// CHECK19-NEXT:    store double* [[A]], double** [[TMP9]], align 4
// CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP10]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP15]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK19-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK19-NEXT:    br i1 [[TMP25]], 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_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR3:[0-9]+]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
// CHECK19-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 4
// CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
// CHECK19-NEXT:    store double* [[A2]], double** [[TMP29]], align 4
// CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP30]], align 4
// CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
// CHECK19-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
// CHECK19-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK19:       omp_offload.failed6:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK19:       omp_offload.cont7:
// CHECK19-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP35:%.*]] = load double, double* [[A8]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = fptosi double [[TMP35]] to i32
// CHECK19-NEXT:    ret i32 [[CONV]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[TMP4]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
// CHECK19-NEXT:    store i32 [[TMP2]], i32* [[TMP6]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP7]], align 4
// CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
// CHECK19-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK19-NEXT:    br i1 [[TMP12]], 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_l104(i32 [[TMP2]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK19-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 4
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK19-NEXT:    store i32 [[TMP15]], i32* [[TMP19]], align 4
// CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK19-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK19-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK19:       omp_offload.failed6:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP15]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK19:       omp_offload.cont7:
// CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP26]], 1
// CHECK19-NEXT:    ret i32 [[ADD8]]
//
//
// 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:    [[B:%.*]] = alloca i16, align 2
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__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:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
// CHECK19-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
// CHECK19-NEXT:    br i1 [[TMP1]], 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_l88() #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK19:       omp_offload.cont:
// CHECK19-NEXT:    store i16 1, i16* [[B]], align 2
// CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
// CHECK19-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP5]], i16* [[CONV]], align 2
// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK19-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP7]], i16* [[CONV1]], align 2
// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK19-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
// CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK19-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
// CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK19-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[TMP15]], align 4
// CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK19-NEXT:    store i32 [[TMP6]], i32* [[TMP17]], align 4
// CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK19-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[TMP20]], align 4
// CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK19-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK19-NEXT:    store i32 [[TMP8]], i32* [[TMP22]], align 4
// CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK19-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK19-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK19-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
// CHECK19-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
// CHECK19-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK19-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
// CHECK19:       omp_offload.failed2:
// CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR3]]
// CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
// CHECK19:       omp_offload.cont3:
// CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
// CHECK19-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
// CHECK19-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]]) #[[ATTR2:[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK19-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] 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:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK19-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[ATTR2]] {
// 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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK19-NEXT:    store double 2.500000e+00, double* [[A]], align 4
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK19-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..4 to void (i32*, i32*, ...)*))
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
// CHECK19-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..7 to void (i32*, i32*, ...)*))
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK19-SAME: () #[[ATTR1]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// 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..10 to void (i32*, i32*, ...)*))
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
// CHECK19-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK19-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
// CHECK19-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK19-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK19-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK19-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[B:%.*]]) #[[ATTR2]] {
// 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 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 [[B]], i32* [[B_ADDR]], align 4
// CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK19-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK19-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
// CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK19-NEXT:    ret void
//
//
// CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK19-SAME: () #[[ATTR4:[0-9]+]] {
// CHECK19-NEXT:  entry:
// CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK19-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@_Z3bari
// 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:    [[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 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], 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 @_ZL7fstatici(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 @_Z9ftemplateIiET_i(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* [[A]], align 4
// CHECK20-NEXT:    ret i32 [[TMP6]]
//
//
// 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:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], 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:    store i32 1, i32* [[B]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
// CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
// CHECK20-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
// CHECK20-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
// CHECK20-NEXT:    store double* [[A]], double** [[TMP9]], align 4
// CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP10]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP15]], align 4
// CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
// CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
// CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK20-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK20-NEXT:    br i1 [[TMP25]], 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_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR3:[0-9]+]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
// CHECK20-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 4
// CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
// CHECK20-NEXT:    store double* [[A2]], double** [[TMP29]], align 4
// CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP30]], align 4
// CHECK20-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
// CHECK20-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
// CHECK20-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK20:       omp_offload.failed6:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK20:       omp_offload.cont7:
// CHECK20-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP35:%.*]] = load double, double* [[A8]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = fptosi double [[TMP35]] to i32
// CHECK20-NEXT:    ret i32 [[CONV]]
//
//
// CHECK20-LABEL: define {{[^@]+}}@_ZL7fstatici
// CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__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:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
// CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[TMP4]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
// CHECK20-NEXT:    store i32 [[TMP2]], i32* [[TMP6]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP7]], align 4
// CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
// CHECK20-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK20-NEXT:    br i1 [[TMP12]], 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_l104(i32 [[TMP2]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
// CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK20-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 4
// CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
// CHECK20-NEXT:    store i32 [[TMP15]], i32* [[TMP19]], align 4
// CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP20]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
// CHECK20-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK20-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
// CHECK20:       omp_offload.failed6:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP15]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
// CHECK20:       omp_offload.cont7:
// CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
// CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP26]], 1
// CHECK20-NEXT:    ret i32 [[ADD8]]
//
//
// 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:    [[B:%.*]] = alloca i16, align 2
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__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:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
// CHECK20-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
// CHECK20-NEXT:    br i1 [[TMP1]], 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_l88() #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK20:       omp_offload.cont:
// CHECK20-NEXT:    store i16 1, i16* [[B]], align 2
// CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
// CHECK20-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP5]], i16* [[CONV]], align 2
// CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK20-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP7]], i16* [[CONV1]], align 2
// CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
// CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
// CHECK20-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
// CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
// CHECK20-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
// CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK20-NEXT:    store i8* null, i8** [[TMP13]], align 4
// CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[TMP15]], align 4
// CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
// CHECK20-NEXT:    store i32 [[TMP6]], i32* [[TMP17]], align 4
// CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK20-NEXT:    store i8* null, i8** [[TMP18]], align 4
// CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[TMP20]], align 4
// CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK20-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK20-NEXT:    store i32 [[TMP8]], i32* [[TMP22]], align 4
// CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK20-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK20-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
// CHECK20-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
// CHECK20-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
// CHECK20-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK20-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
// CHECK20:       omp_offload.failed2:
// CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR3]]
// CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
// CHECK20:       omp_offload.cont3:
// CHECK20-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
// CHECK20-NEXT:    ret i32 [[TMP30]]
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
// CHECK20-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[B:%.*]]) #[[ATTR2:[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK20-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] 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:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK20-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[ATTR2]] {
// 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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK20-NEXT:    store double 2.500000e+00, double* [[A]], align 4
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK20-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..4 to void (i32*, i32*, ...)*))
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
// CHECK20-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// 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..7 to void (i32*, i32*, ...)*))
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK20-SAME: () #[[ATTR1]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// 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..10 to void (i32*, i32*, ...)*))
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..10
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// 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_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
// CHECK20-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
// CHECK20-NEXT:  entry:
// CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK20-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
// CHECK20-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK20-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK20-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK20-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[B:%.*]]) #[[ATTR2]] {
// 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 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 [[B]], i32* [[B_ADDR]], align 4
// CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK20-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK20-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
// CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK20-NEXT:    ret void
//
//
// CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK20-SAME: () #[[ATTR4:[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]+}}__ZL7fstatici_l104
// CHECK25-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK25-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*, ...)*))
// 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]+}}__ZL7fstatici_l108
// CHECK25-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK25-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..1 to void (i32*, i32*, ...)*))
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__ZN2S12r1Ei_l121
// CHECK25-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
// CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK25-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK25-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK25-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[B]], i64* [[B_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 i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
// CHECK25-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 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:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK25-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK25-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK25-NEXT:    store double 2.500000e+00, double* [[A]], align 8
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK25-SAME: () #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// CHECK25-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..4 to void (i32*, i32*, ...)*))
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__Z9ftemplateIiET_i_l93
// CHECK25-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK25-NEXT:  entry:
// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK25-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
// CHECK25-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK25-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
// CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK25-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK25-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
// CHECK25-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
// CHECK25-NEXT:    ret void
//
//
// CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[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:    [[B_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 [[B]], i64* [[B_ADDR]], align 8
// CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK25-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK25-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
// CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
// CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK25-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK26-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK26-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*, ...)*))
// 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]+}}__ZL7fstatici_l108
// CHECK26-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK26-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..1 to void (i32*, i32*, ...)*))
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__ZN2S12r1Ei_l121
// CHECK26-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
// CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
// CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
// CHECK26-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
// CHECK26-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i64 [[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK26-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[B]], i64* [[B_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 i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
// CHECK26-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 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:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK26-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
// CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
// CHECK26-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK26-NEXT:    store double 2.500000e+00, double* [[A]], align 8
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK26-SAME: () #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// CHECK26-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..4 to void (i32*, i32*, ...)*))
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__Z9ftemplateIiET_i_l93
// CHECK26-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK26-NEXT:  entry:
// CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK26-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
// CHECK26-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
// CHECK26-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
// CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
// CHECK26-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK26-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
// CHECK26-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
// CHECK26-NEXT:    ret void
//
//
// CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[A:%.*]], i64 [[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:    [[B_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 [[B]], i64* [[B_ADDR]], align 8
// CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
// CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK26-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 8
// CHECK26-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
// CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
// CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
// CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 8
// CHECK26-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK27-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK27-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*, ...)*))
// 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]+}}__ZL7fstatici_l108
// CHECK27-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK27-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..1 to void (i32*, i32*, ...)*))
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__ZN2S12r1Ei_l121
// CHECK27-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
// CHECK27-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK27-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] 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:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK27-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK27-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK27-NEXT:    store double 2.500000e+00, double* [[A]], align 4
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK27-SAME: () #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// CHECK27-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..4 to void (i32*, i32*, ...)*))
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__Z9ftemplateIiET_i_l93
// CHECK27-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK27-NEXT:  entry:
// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK27-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
// CHECK27-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK27-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK27-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK27-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
// CHECK27-NEXT:    ret void
//
//
// CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[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:    [[B_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 [[B]], i32* [[B_ADDR]], align 4
// CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK27-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK27-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
// CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK27-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
// CHECK28-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
// CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK28-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*, ...)*))
// 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]+}}__ZL7fstatici_l108
// CHECK28-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
// CHECK28-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..1 to void (i32*, i32*, ...)*))
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__ZN2S12r1Ei_l121
// CHECK28-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// 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 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
// CHECK28-NEXT:    [[TMP4:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]], i32 [[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:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK28-NEXT:    [[B_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 %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] 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:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
// CHECK28-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
// 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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S1* [[THIS:%.*]]) #[[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:    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:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
// CHECK28-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
// CHECK28-NEXT:    store double 2.500000e+00, double* [[A]], align 4
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
// CHECK28-SAME: () #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
// CHECK28-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..4 to void (i32*, i32*, ...)*))
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..4
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
// 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]+}}__Z9ftemplateIiET_i_l93
// CHECK28-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
// CHECK28-NEXT:  entry:
// CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
// CHECK28-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
// CHECK28-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
// CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
// CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
// CHECK28-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK28-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
// CHECK28-NEXT:    [[TMP6:%.*]] = 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 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
// CHECK28-NEXT:    ret void
//
//
// CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[A:%.*]], i32 [[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:    [[B_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 [[B]], i32* [[B_ADDR]], align 4
// CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK28-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK28-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
// CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
// CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
// CHECK28-NEXT:    ret void
//