// 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 _
// 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=CHECK1
// 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=CHECK2
// 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=CHECK3
// 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=CHECK4

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


double Ga = 1.0;
double Gb = 2.0;
double Gc = 3.0;
double Gd = 4.0;

int foo(short a, short b, short c, short d){
  static float Sa = 5.0;
  static float Sb = 6.0;
  static float Sc = 7.0;
  static float Sd = 8.0;


  // 3 local vars being captured.


  // 3 static vars being captured.


  // 3 static global vars being captured.


  // Capture b, Gb, Sb, Gc, c, Sc, d, Gd, Sd
  #pragma omp target if(Ga>0.0 && a>0 && Sa>0.0)
  {
    b += 1;
    Gb += 1.0;
    Sb += 1.0;

    // The parallel region only uses 3 captures.
    // Capture d, Gd, Sd,

    #pragma omp parallel if(Gc>0.0 && c>0 && Sc>0.0)
    {
      d += 1;
      Gd += 1.0;
      Sd += 1.0;
    }
  }
  return a + b + c + d + (int)Sa + (int)Sb + (int)Sc + (int)Sd;
}

int bar(short a, short b, short c, short d){
  static float Sa = 9.0;
  static float Sb = 10.0;
  static float Sc = 11.0;
  static float Sd = 12.0;

  // Capture a, b, c, d
  #pragma omp parallel
  {

    // 3 local vars being captured.


    // 3 static vars being captured.


    // 3 static global vars being captured.


    // Capture b, Gb, Sb, Gc, c, Sc, d, Gd, Sd
    #pragma omp target if(Ga>0.0 && a>0 && Sa>0.0)
    {
      b += 1;
      Gb += 1.0;
      Sb += 1.0;


      // Capture d, Gd, Sd
      #pragma omp parallel if(Gc>0.0 && c>0 && Sc>0.0)
      {
        d += 1;
        Gd += 1.0;
        Sd += 1.0;
      }
    }
  }
  return a + b + c + d + (int)Sa + (int)Sb + (int)Sc + (int)Sd;
}

///
/// Tests with template functions.
///


template<typename T>
int tbar(T a, T b, T c, T d){
  static float Sa = 17.0;
  static float Sb = 18.0;
  static float Sc = 19.0;
  static float Sd = 20.0;

  // Capture a, b, c, d
  #pragma omp parallel
  {

    // 3 local vars being captured.


    // 3 static vars being captured.


    // 3 static global vars being captured.


    // Capture b, Gb, Sb, Gc, c, Sc, d, Gd, Sd
    #pragma omp target if(Ga>0.0 && a>0 && Sa>0.0)
    {
      b += 1;
      Gb += 1.0;
      Sb += 1.0;


      // Capture d, Gd, Sd
      #pragma omp parallel if(Gc>0.0 && c>0 && Sc>0.0)
      {
        d += 1;
        Gd += 1.0;
        Sd += 1.0;
      }
    }
  }
  return a + b + c + d + (int)Sa + (int)Sb + (int)Sc + (int)Sd;
}

int tbar2(short a, short b, short c, short d){
  return tbar(a, b, c, d);
}

#endif
// CHECK1-LABEL: define {{[^@]+}}@_Z3foossss
// CHECK1-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GB_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SB_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GC_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SC_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[D_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GD_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SD_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP0]], i16* [[CONV]], align 2
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load double, double* @Gb, align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP2]], double* [[CONV1]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[GB_CASTED]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ3foossssE2Sb, align 4
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP4]], float* [[CONV2]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[SB_CASTED]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load double, double* @Gc, align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP6]], double* [[CONV3]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[GC_CASTED]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP8]], i16* [[CONV4]], align 2
// CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[C_CASTED]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = load float, float* @_ZZ3foossssE2Sc, align 4
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP10]], float* [[CONV5]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[SC_CASTED]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP12]], i16* [[CONV6]], align 2
// CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[D_CASTED]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = load double, double* @Gd, align 8
// CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP14]], double* [[CONV7]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[GD_CASTED]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = load float, float* @_ZZ3foossssE2Sd, align 4
// CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP16]], float* [[CONV8]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[SD_CASTED]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = load double, double* @Ga, align 8
// CHECK1-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP18]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       land.lhs.true:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP19]] to i32
// CHECK1-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[CONV9]], 0
// CHECK1-NEXT:    br i1 [[CMP10]], label [[LAND_LHS_TRUE11:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       land.lhs.true11:
// CHECK1-NEXT:    [[TMP20:%.*]] = load float, float* @_ZZ3foossssE2Sa, align 4
// CHECK1-NEXT:    [[CONV12:%.*]] = fpext float [[TMP20]] to double
// CHECK1-NEXT:    [[CMP13:%.*]] = fcmp ogt double [[CONV12]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP13]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
// CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP34]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP35]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP37]], align 8
// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP39]], align 8
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK1-NEXT:    store i8* null, i8** [[TMP40]], align 8
// CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i64*
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP42]], align 8
// CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i64*
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP44]], align 8
// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP45]], align 8
// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i64*
// CHECK1-NEXT:    store i64 [[TMP11]], i64* [[TMP47]], align 8
// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i64*
// CHECK1-NEXT:    store i64 [[TMP11]], i64* [[TMP49]], align 8
// CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
// CHECK1-NEXT:    store i8* null, i8** [[TMP50]], align 8
// CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
// CHECK1-NEXT:    store i64 [[TMP13]], i64* [[TMP52]], align 8
// CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP54:%.*]] = bitcast i8** [[TMP53]] to i64*
// CHECK1-NEXT:    store i64 [[TMP13]], i64* [[TMP54]], align 8
// CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 6
// CHECK1-NEXT:    store i8* null, i8** [[TMP55]], align 8
// CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i64*
// CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP57]], align 8
// CHECK1-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i64*
// CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP59]], align 8
// CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 7
// CHECK1-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
// CHECK1-NEXT:    store i64 [[TMP17]], i64* [[TMP62]], align 8
// CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
// CHECK1-NEXT:    store i64 [[TMP17]], i64* [[TMP64]], align 8
// CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 8
// CHECK1-NEXT:    store i8* null, i8** [[TMP65]], align 8
// CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49.region_id, i32 9, i8** [[TMP66]], i8** [[TMP67]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null)
// CHECK1-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
// CHECK1-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]]) #[[ATTR2:[0-9]+]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    [[TMP70:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP70]] to i32
// CHECK1-NEXT:    [[TMP71:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP71]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV14]], [[CONV15]]
// CHECK1-NEXT:    [[TMP72:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    [[CONV16:%.*]] = sext i16 [[TMP72]] to i32
// CHECK1-NEXT:    [[ADD17:%.*]] = add nsw i32 [[ADD]], [[CONV16]]
// CHECK1-NEXT:    [[TMP73:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    [[CONV18:%.*]] = sext i16 [[TMP73]] to i32
// CHECK1-NEXT:    [[ADD19:%.*]] = add nsw i32 [[ADD17]], [[CONV18]]
// CHECK1-NEXT:    [[TMP74:%.*]] = load float, float* @_ZZ3foossssE2Sa, align 4
// CHECK1-NEXT:    [[CONV20:%.*]] = fptosi float [[TMP74]] to i32
// CHECK1-NEXT:    [[ADD21:%.*]] = add nsw i32 [[ADD19]], [[CONV20]]
// CHECK1-NEXT:    [[TMP75:%.*]] = load float, float* @_ZZ3foossssE2Sb, align 4
// CHECK1-NEXT:    [[CONV22:%.*]] = fptosi float [[TMP75]] to i32
// CHECK1-NEXT:    [[ADD23:%.*]] = add nsw i32 [[ADD21]], [[CONV22]]
// CHECK1-NEXT:    [[TMP76:%.*]] = load float, float* @_ZZ3foossssE2Sc, align 4
// CHECK1-NEXT:    [[CONV24:%.*]] = fptosi float [[TMP76]] to i32
// CHECK1-NEXT:    [[ADD25:%.*]] = add nsw i32 [[ADD23]], [[CONV24]]
// CHECK1-NEXT:    [[TMP77:%.*]] = load float, float* @_ZZ3foossssE2Sd, align 4
// CHECK1-NEXT:    [[CONV26:%.*]] = fptosi float [[TMP77]] to i32
// CHECK1-NEXT:    [[ADD27:%.*]] = add nsw i32 [[ADD25]], [[CONV26]]
// CHECK1-NEXT:    ret i32 [[ADD27]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49
// CHECK1-SAME: (i64 [[B:%.*]], i64 [[GB:%.*]], i64 [[SB:%.*]], i64 [[GC:%.*]], i64 [[C:%.*]], i64 [[SC:%.*]], i64 [[D:%.*]], i64 [[GD:%.*]], i64 [[SD:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GB_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SB_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GC_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SC_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GD_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SD_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GB]], i64* [[GB_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SB]], i64* [[SB_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GC]], i64* [[GC_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[C]], i64* [[C_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SC]], i64* [[SC_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[D]], i64* [[D_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GD]], i64* [[GD_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SD]], i64* [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_ADDR]] to double*
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_ADDR]] to float*
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_ADDR]] to double*
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_ADDR]] to float*
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_ADDR]] to double*
// CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_ADDR]] to float*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP1]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK1-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK1-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load double, double* [[CONV1]], align 8
// CHECK1-NEXT:    [[ADD11:%.*]] = fadd double [[TMP2]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD11]], double* [[CONV1]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load float, float* [[CONV2]], align 8
// CHECK1-NEXT:    [[CONV12:%.*]] = fpext float [[TMP3]] to double
// CHECK1-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK1-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK1-NEXT:    store float [[CONV14]], float* [[CONV2]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load double, double* [[CONV3]], align 8
// CHECK1-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP4]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       land.lhs.true:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV4]], align 8
// CHECK1-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP5]] to i32
// CHECK1-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK1-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       land.lhs.true17:
// CHECK1-NEXT:    [[TMP6:%.*]] = load float, float* [[CONV5]], align 8
// CHECK1-NEXT:    [[CONV18:%.*]] = fpext float [[TMP6]] to double
// CHECK1-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined. to void (i32*, i32*, ...)*), i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]])
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK1-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]]) #[[ATTR2]]
// CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 8 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 8
// CHECK1-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK1-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK1-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK1-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK1-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK1-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK1-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK1-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z3barssss
// CHECK1-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, i16*, i16*, i16*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i16* [[A_ADDR]], i16* [[B_ADDR]], i16* [[C_ADDR]], i16* [[D_ADDR]])
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV1]]
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CONV2]]
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP3]] to i32
// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CONV4]]
// CHECK1-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ3barssssE2Sa, align 4
// CHECK1-NEXT:    [[CONV6:%.*]] = fptosi float [[TMP4]] to i32
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CONV6]]
// CHECK1-NEXT:    [[TMP5:%.*]] = load float, float* @_ZZ3barssssE2Sb, align 4
// CHECK1-NEXT:    [[CONV8:%.*]] = fptosi float [[TMP5]] to i32
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD7]], [[CONV8]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ3barssssE2Sc, align 4
// CHECK1-NEXT:    [[CONV10:%.*]] = fptosi float [[TMP6]] to i32
// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CONV10]]
// CHECK1-NEXT:    [[TMP7:%.*]] = load float, float* @_ZZ3barssssE2Sd, align 4
// CHECK1-NEXT:    [[CONV12:%.*]] = fptosi float [[TMP7]] to i32
// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[ADD11]], [[CONV12]]
// CHECK1-NEXT:    ret i32 [[ADD13]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]]) #[[ATTR1]] {
// 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 i16*, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GB_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SB_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GC_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SC_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[D_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GD_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SD_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], 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 i16* [[A]], i16** [[A_ADDR]], align 8
// CHECK1-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 8
// CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK1-NEXT:    store i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP1]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load double, double* @Gb, align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP6]], double* [[CONV1]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[GB_CASTED]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load float, float* @_ZZ3barssssE2Sb, align 4
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP8]], float* [[CONV2]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[SB_CASTED]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = load double, double* @Gc, align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP10]], double* [[CONV3]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[GC_CASTED]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP2]], align 2
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP12]], i16* [[CONV4]], align 2
// CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[C_CASTED]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ3barssssE2Sc, align 4
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP14]], float* [[CONV5]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[SC_CASTED]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP3]], align 2
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP16]], i16* [[CONV6]], align 2
// CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[D_CASTED]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = load double, double* @Gd, align 8
// CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP18]], double* [[CONV7]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[GD_CASTED]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = load float, float* @_ZZ3barssssE2Sd, align 4
// CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP20]], float* [[CONV8]], align 4
// CHECK1-NEXT:    [[TMP21:%.*]] = load i64, i64* [[SD_CASTED]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = load double, double* @Ga, align 8
// CHECK1-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP22]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       land.lhs.true:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK1-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP23]] to i32
// CHECK1-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[CONV9]], 0
// CHECK1-NEXT:    br i1 [[CMP10]], label [[LAND_LHS_TRUE11:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       land.lhs.true11:
// CHECK1-NEXT:    [[TMP24:%.*]] = load float, float* @_ZZ3barssssE2Sa, align 4
// CHECK1-NEXT:    [[CONV12:%.*]] = fpext float [[TMP24]] to double
// CHECK1-NEXT:    [[CMP13:%.*]] = fcmp ogt double [[CONV12]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP13]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i64*
// CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP34]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i64*
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP36]], align 8
// CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i64*
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP39]], align 8
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
// CHECK1-NEXT:    store i64 [[TMP11]], i64* [[TMP41]], align 8
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK1-NEXT:    store i64 [[TMP11]], i64* [[TMP43]], align 8
// CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK1-NEXT:    store i8* null, i8** [[TMP44]], align 8
// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK1-NEXT:    store i64 [[TMP13]], i64* [[TMP46]], align 8
// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK1-NEXT:    store i64 [[TMP13]], i64* [[TMP48]], align 8
// CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
// CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP51]], align 8
// CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
// CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP53]], align 8
// CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
// CHECK1-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i64*
// CHECK1-NEXT:    store i64 [[TMP17]], i64* [[TMP56]], align 8
// CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i64*
// CHECK1-NEXT:    store i64 [[TMP17]], i64* [[TMP58]], align 8
// CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 6
// CHECK1-NEXT:    store i8* null, i8** [[TMP59]], align 8
// CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP61:%.*]] = bitcast i8** [[TMP60]] to i64*
// CHECK1-NEXT:    store i64 [[TMP19]], i64* [[TMP61]], align 8
// CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP63:%.*]] = bitcast i8** [[TMP62]] to i64*
// CHECK1-NEXT:    store i64 [[TMP19]], i64* [[TMP63]], align 8
// CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 7
// CHECK1-NEXT:    store i8* null, i8** [[TMP64]], align 8
// CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP66:%.*]] = bitcast i8** [[TMP65]] to i64*
// CHECK1-NEXT:    store i64 [[TMP21]], i64* [[TMP66]], align 8
// CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to i64*
// CHECK1-NEXT:    store i64 [[TMP21]], i64* [[TMP68]], align 8
// CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 8
// CHECK1-NEXT:    store i8* null, i8** [[TMP69]], align 8
// CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP72:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94.region_id, i32 9, i8** [[TMP70]], i8** [[TMP71]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes.3, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null)
// CHECK1-NEXT:    [[TMP73:%.*]] = icmp ne i32 [[TMP72]], 0
// CHECK1-NEXT:    br i1 [[TMP73]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94(i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94(i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94
// CHECK1-SAME: (i64 [[B:%.*]], i64 [[GB:%.*]], i64 [[SB:%.*]], i64 [[GC:%.*]], i64 [[C:%.*]], i64 [[SC:%.*]], i64 [[D:%.*]], i64 [[GD:%.*]], i64 [[SD:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GB_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SB_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GC_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SC_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GD_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SD_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GB]], i64* [[GB_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SB]], i64* [[SB_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GC]], i64* [[GC_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[C]], i64* [[C_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SC]], i64* [[SC_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[D]], i64* [[D_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GD]], i64* [[GD_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SD]], i64* [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_ADDR]] to double*
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_ADDR]] to float*
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_ADDR]] to double*
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_ADDR]] to float*
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_ADDR]] to double*
// CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_ADDR]] to float*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP1]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK1-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK1-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load double, double* [[CONV1]], align 8
// CHECK1-NEXT:    [[ADD11:%.*]] = fadd double [[TMP2]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD11]], double* [[CONV1]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load float, float* [[CONV2]], align 8
// CHECK1-NEXT:    [[CONV12:%.*]] = fpext float [[TMP3]] to double
// CHECK1-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK1-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK1-NEXT:    store float [[CONV14]], float* [[CONV2]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load double, double* [[CONV3]], align 8
// CHECK1-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP4]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       land.lhs.true:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV4]], align 8
// CHECK1-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP5]] to i32
// CHECK1-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK1-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       land.lhs.true17:
// CHECK1-NEXT:    [[TMP6:%.*]] = load float, float* [[CONV5]], align 8
// CHECK1-NEXT:    [[CONV18:%.*]] = fpext float [[TMP6]] to double
// CHECK1-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]])
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK1-NEXT:    call void @.omp_outlined..2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]]) #[[ATTR2]]
// CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 8 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 8
// CHECK1-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK1-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK1-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK1-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK1-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK1-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK1-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK1-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z5tbar2ssss
// CHECK1-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    [[CALL:%.*]] = call signext i32 @_Z4tbarIsEiT_S0_S0_S0_(i16 signext [[TMP0]], i16 signext [[TMP1]], i16 signext [[TMP2]], i16 signext [[TMP3]])
// CHECK1-NEXT:    ret i32 [[CALL]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z4tbarIsEiT_S0_S0_S0_
// CHECK1-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] comdat {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK1-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, i16*, i16*, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i16* [[A_ADDR]], i16* [[B_ADDR]], i16* [[C_ADDR]], i16* [[D_ADDR]])
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK1-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV1]]
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CONV2]]
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK1-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP3]] to i32
// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CONV4]]
// CHECK1-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sa, align 4
// CHECK1-NEXT:    [[CONV6:%.*]] = fptosi float [[TMP4]] to i32
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CONV6]]
// CHECK1-NEXT:    [[TMP5:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sb, align 4
// CHECK1-NEXT:    [[CONV8:%.*]] = fptosi float [[TMP5]] to i32
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD7]], [[CONV8]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sc, align 4
// CHECK1-NEXT:    [[CONV10:%.*]] = fptosi float [[TMP6]] to i32
// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CONV10]]
// CHECK1-NEXT:    [[TMP7:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sd, align 4
// CHECK1-NEXT:    [[CONV12:%.*]] = fptosi float [[TMP7]] to i32
// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[ADD11]], [[CONV12]]
// CHECK1-NEXT:    ret i32 [[ADD13]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]]) #[[ATTR1]] {
// 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 i16*, align 8
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GB_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SB_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GC_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SC_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[D_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GD_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SD_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], 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 i16* [[A]], i16** [[A_ADDR]], align 8
// CHECK1-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 8
// CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK1-NEXT:    store i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP1]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load double, double* @Gb, align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP6]], double* [[CONV1]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[GB_CASTED]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sb, align 4
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP8]], float* [[CONV2]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[SB_CASTED]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = load double, double* @Gc, align 8
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP10]], double* [[CONV3]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[GC_CASTED]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP2]], align 2
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP12]], i16* [[CONV4]], align 2
// CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[C_CASTED]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sc, align 4
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP14]], float* [[CONV5]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[SC_CASTED]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP3]], align 2
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_CASTED]] to i16*
// CHECK1-NEXT:    store i16 [[TMP16]], i16* [[CONV6]], align 2
// CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[D_CASTED]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = load double, double* @Gd, align 8
// CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_CASTED]] to double*
// CHECK1-NEXT:    store double [[TMP18]], double* [[CONV7]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[GD_CASTED]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sd, align 4
// CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_CASTED]] to float*
// CHECK1-NEXT:    store float [[TMP20]], float* [[CONV8]], align 4
// CHECK1-NEXT:    [[TMP21:%.*]] = load i64, i64* [[SD_CASTED]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = load double, double* @Ga, align 8
// CHECK1-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP22]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       land.lhs.true:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK1-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP23]] to i32
// CHECK1-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[CONV9]], 0
// CHECK1-NEXT:    br i1 [[CMP10]], label [[LAND_LHS_TRUE11:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       land.lhs.true11:
// CHECK1-NEXT:    [[TMP24:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sa, align 4
// CHECK1-NEXT:    [[CONV12:%.*]] = fpext float [[TMP24]] to double
// CHECK1-NEXT:    [[CMP13:%.*]] = fcmp ogt double [[CONV12]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP13]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
// CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i64*
// CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK1-NEXT:    store i8* null, i8** [[TMP34]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i64*
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP36]], align 8
// CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i64*
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK1-NEXT:    store i8* null, i8** [[TMP39]], align 8
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
// CHECK1-NEXT:    store i64 [[TMP11]], i64* [[TMP41]], align 8
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK1-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK1-NEXT:    store i64 [[TMP11]], i64* [[TMP43]], align 8
// CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK1-NEXT:    store i8* null, i8** [[TMP44]], align 8
// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK1-NEXT:    store i64 [[TMP13]], i64* [[TMP46]], align 8
// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK1-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK1-NEXT:    store i64 [[TMP13]], i64* [[TMP48]], align 8
// CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK1-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
// CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP51]], align 8
// CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK1-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
// CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP53]], align 8
// CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
// CHECK1-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i64*
// CHECK1-NEXT:    store i64 [[TMP17]], i64* [[TMP56]], align 8
// CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK1-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i64*
// CHECK1-NEXT:    store i64 [[TMP17]], i64* [[TMP58]], align 8
// CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 6
// CHECK1-NEXT:    store i8* null, i8** [[TMP59]], align 8
// CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP61:%.*]] = bitcast i8** [[TMP60]] to i64*
// CHECK1-NEXT:    store i64 [[TMP19]], i64* [[TMP61]], align 8
// CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK1-NEXT:    [[TMP63:%.*]] = bitcast i8** [[TMP62]] to i64*
// CHECK1-NEXT:    store i64 [[TMP19]], i64* [[TMP63]], align 8
// CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 7
// CHECK1-NEXT:    store i8* null, i8** [[TMP64]], align 8
// CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP66:%.*]] = bitcast i8** [[TMP65]] to i64*
// CHECK1-NEXT:    store i64 [[TMP21]], i64* [[TMP66]], align 8
// CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK1-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to i64*
// CHECK1-NEXT:    store i64 [[TMP21]], i64* [[TMP68]], align 8
// CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 8
// CHECK1-NEXT:    store i8* null, i8** [[TMP69]], align 8
// CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP72:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145.region_id, i32 9, i8** [[TMP70]], i8** [[TMP71]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null)
// CHECK1-NEXT:    [[TMP73:%.*]] = icmp ne i32 [[TMP72]], 0
// CHECK1-NEXT:    br i1 [[TMP73]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145(i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145(i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145
// CHECK1-SAME: (i64 [[B:%.*]], i64 [[GB:%.*]], i64 [[SB:%.*]], i64 [[GC:%.*]], i64 [[C:%.*]], i64 [[SC:%.*]], i64 [[D:%.*]], i64 [[GD:%.*]], i64 [[SD:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GB_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SB_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GC_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SC_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[GD_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[SD_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GB]], i64* [[GB_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SB]], i64* [[SB_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GC]], i64* [[GC_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[C]], i64* [[C_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SC]], i64* [[SC_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[D]], i64* [[D_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[GD]], i64* [[GD_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[SD]], i64* [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_ADDR]] to double*
// CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_ADDR]] to float*
// CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_ADDR]] to double*
// CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_ADDR]] to float*
// CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_ADDR]] to i16*
// CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_ADDR]] to double*
// CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_ADDR]] to float*
// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK1-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP1]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK1-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK1-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load double, double* [[CONV1]], align 8
// CHECK1-NEXT:    [[ADD11:%.*]] = fadd double [[TMP2]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD11]], double* [[CONV1]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load float, float* [[CONV2]], align 8
// CHECK1-NEXT:    [[CONV12:%.*]] = fpext float [[TMP3]] to double
// CHECK1-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK1-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK1-NEXT:    store float [[CONV14]], float* [[CONV2]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load double, double* [[CONV3]], align 8
// CHECK1-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP4]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK1:       land.lhs.true:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV4]], align 8
// CHECK1-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP5]] to i32
// CHECK1-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK1-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       land.lhs.true17:
// CHECK1-NEXT:    [[TMP6:%.*]] = load float, float* [[CONV5]], align 8
// CHECK1-NEXT:    [[CONV18:%.*]] = fpext float [[TMP6]] to double
// CHECK1-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK1:       omp_if.then:
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]])
// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK1:       omp_if.else:
// CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK1-NEXT:    call void @.omp_outlined..6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]]) #[[ATTR2]]
// CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK1-NEXT:    br label [[OMP_IF_END]]
// CHECK1:       omp_if.end:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 8 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK1-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 8
// CHECK1-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 8
// CHECK1-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK1-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK1-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK1-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK1-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK1-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK1-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK1-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK1-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z3foossss
// CHECK2-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GB_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SB_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GC_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SC_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[D_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GD_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SD_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP0]], i16* [[CONV]], align 2
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load double, double* @Gb, align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP2]], double* [[CONV1]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[GB_CASTED]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ3foossssE2Sb, align 4
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP4]], float* [[CONV2]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[SB_CASTED]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load double, double* @Gc, align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP6]], double* [[CONV3]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[GC_CASTED]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP8]], i16* [[CONV4]], align 2
// CHECK2-NEXT:    [[TMP9:%.*]] = load i64, i64* [[C_CASTED]], align 8
// CHECK2-NEXT:    [[TMP10:%.*]] = load float, float* @_ZZ3foossssE2Sc, align 4
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP10]], float* [[CONV5]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i64, i64* [[SC_CASTED]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP12]], i16* [[CONV6]], align 2
// CHECK2-NEXT:    [[TMP13:%.*]] = load i64, i64* [[D_CASTED]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = load double, double* @Gd, align 8
// CHECK2-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP14]], double* [[CONV7]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[GD_CASTED]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = load float, float* @_ZZ3foossssE2Sd, align 4
// CHECK2-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP16]], float* [[CONV8]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[SD_CASTED]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = load double, double* @Ga, align 8
// CHECK2-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP18]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       land.lhs.true:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP19]] to i32
// CHECK2-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[CONV9]], 0
// CHECK2-NEXT:    br i1 [[CMP10]], label [[LAND_LHS_TRUE11:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       land.lhs.true11:
// CHECK2-NEXT:    [[TMP20:%.*]] = load float, float* @_ZZ3foossssE2Sa, align 4
// CHECK2-NEXT:    [[CONV12:%.*]] = fpext float [[TMP20]] to double
// CHECK2-NEXT:    [[CMP13:%.*]] = fcmp ogt double [[CONV12]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP13]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
// CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP22]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
// CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP24]], align 8
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP29]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP30]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP32]], align 8
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP34]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP35]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
// CHECK2-NEXT:    store i64 [[TMP7]], i64* [[TMP37]], align 8
// CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
// CHECK2-NEXT:    store i64 [[TMP7]], i64* [[TMP39]], align 8
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK2-NEXT:    store i8* null, i8** [[TMP40]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i64*
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP42]], align 8
// CHECK2-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i64*
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP44]], align 8
// CHECK2-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK2-NEXT:    store i8* null, i8** [[TMP45]], align 8
// CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK2-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i64*
// CHECK2-NEXT:    store i64 [[TMP11]], i64* [[TMP47]], align 8
// CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK2-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i64*
// CHECK2-NEXT:    store i64 [[TMP11]], i64* [[TMP49]], align 8
// CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
// CHECK2-NEXT:    store i8* null, i8** [[TMP50]], align 8
// CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK2-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
// CHECK2-NEXT:    store i64 [[TMP13]], i64* [[TMP52]], align 8
// CHECK2-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK2-NEXT:    [[TMP54:%.*]] = bitcast i8** [[TMP53]] to i64*
// CHECK2-NEXT:    store i64 [[TMP13]], i64* [[TMP54]], align 8
// CHECK2-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 6
// CHECK2-NEXT:    store i8* null, i8** [[TMP55]], align 8
// CHECK2-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK2-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i64*
// CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP57]], align 8
// CHECK2-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK2-NEXT:    [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i64*
// CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP59]], align 8
// CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 7
// CHECK2-NEXT:    store i8* null, i8** [[TMP60]], align 8
// CHECK2-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK2-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
// CHECK2-NEXT:    store i64 [[TMP17]], i64* [[TMP62]], align 8
// CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK2-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
// CHECK2-NEXT:    store i64 [[TMP17]], i64* [[TMP64]], align 8
// CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 8
// CHECK2-NEXT:    store i8* null, i8** [[TMP65]], align 8
// CHECK2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49.region_id, i32 9, i8** [[TMP66]], i8** [[TMP67]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null)
// CHECK2-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
// CHECK2-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]]) #[[ATTR2:[0-9]+]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    [[TMP70:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP70]] to i32
// CHECK2-NEXT:    [[TMP71:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP71]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV14]], [[CONV15]]
// CHECK2-NEXT:    [[TMP72:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    [[CONV16:%.*]] = sext i16 [[TMP72]] to i32
// CHECK2-NEXT:    [[ADD17:%.*]] = add nsw i32 [[ADD]], [[CONV16]]
// CHECK2-NEXT:    [[TMP73:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    [[CONV18:%.*]] = sext i16 [[TMP73]] to i32
// CHECK2-NEXT:    [[ADD19:%.*]] = add nsw i32 [[ADD17]], [[CONV18]]
// CHECK2-NEXT:    [[TMP74:%.*]] = load float, float* @_ZZ3foossssE2Sa, align 4
// CHECK2-NEXT:    [[CONV20:%.*]] = fptosi float [[TMP74]] to i32
// CHECK2-NEXT:    [[ADD21:%.*]] = add nsw i32 [[ADD19]], [[CONV20]]
// CHECK2-NEXT:    [[TMP75:%.*]] = load float, float* @_ZZ3foossssE2Sb, align 4
// CHECK2-NEXT:    [[CONV22:%.*]] = fptosi float [[TMP75]] to i32
// CHECK2-NEXT:    [[ADD23:%.*]] = add nsw i32 [[ADD21]], [[CONV22]]
// CHECK2-NEXT:    [[TMP76:%.*]] = load float, float* @_ZZ3foossssE2Sc, align 4
// CHECK2-NEXT:    [[CONV24:%.*]] = fptosi float [[TMP76]] to i32
// CHECK2-NEXT:    [[ADD25:%.*]] = add nsw i32 [[ADD23]], [[CONV24]]
// CHECK2-NEXT:    [[TMP77:%.*]] = load float, float* @_ZZ3foossssE2Sd, align 4
// CHECK2-NEXT:    [[CONV26:%.*]] = fptosi float [[TMP77]] to i32
// CHECK2-NEXT:    [[ADD27:%.*]] = add nsw i32 [[ADD25]], [[CONV26]]
// CHECK2-NEXT:    ret i32 [[ADD27]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49
// CHECK2-SAME: (i64 [[B:%.*]], i64 [[GB:%.*]], i64 [[SB:%.*]], i64 [[GC:%.*]], i64 [[C:%.*]], i64 [[SC:%.*]], i64 [[D:%.*]], i64 [[GD:%.*]], i64 [[SD:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GB_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SB_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GC_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SC_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GD_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SD_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GB]], i64* [[GB_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SB]], i64* [[SB_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GC]], i64* [[GC_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[C]], i64* [[C_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SC]], i64* [[SC_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[D]], i64* [[D_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GD]], i64* [[GD_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SD]], i64* [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_ADDR]] to double*
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_ADDR]] to float*
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_ADDR]] to double*
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_ADDR]] to float*
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_ADDR]] to double*
// CHECK2-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_ADDR]] to float*
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP1]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK2-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK2-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load double, double* [[CONV1]], align 8
// CHECK2-NEXT:    [[ADD11:%.*]] = fadd double [[TMP2]], 1.000000e+00
// CHECK2-NEXT:    store double [[ADD11]], double* [[CONV1]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load float, float* [[CONV2]], align 8
// CHECK2-NEXT:    [[CONV12:%.*]] = fpext float [[TMP3]] to double
// CHECK2-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK2-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK2-NEXT:    store float [[CONV14]], float* [[CONV2]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load double, double* [[CONV3]], align 8
// CHECK2-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP4]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       land.lhs.true:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV4]], align 8
// CHECK2-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP5]] to i32
// CHECK2-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK2-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       land.lhs.true17:
// CHECK2-NEXT:    [[TMP6:%.*]] = load float, float* [[CONV5]], align 8
// CHECK2-NEXT:    [[CONV18:%.*]] = fpext float [[TMP6]] to double
// CHECK2-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined. to void (i32*, i32*, ...)*), i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]])
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK2-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]]) #[[ATTR2]]
// CHECK2-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 8 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 8
// CHECK2-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK2-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK2-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK2-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK2-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK2-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK2-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK2-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK2-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z3barssss
// CHECK2-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, i16*, i16*, i16*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i16* [[A_ADDR]], i16* [[B_ADDR]], i16* [[C_ADDR]], i16* [[D_ADDR]])
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV1]]
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CONV2]]
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP3]] to i32
// CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CONV4]]
// CHECK2-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ3barssssE2Sa, align 4
// CHECK2-NEXT:    [[CONV6:%.*]] = fptosi float [[TMP4]] to i32
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CONV6]]
// CHECK2-NEXT:    [[TMP5:%.*]] = load float, float* @_ZZ3barssssE2Sb, align 4
// CHECK2-NEXT:    [[CONV8:%.*]] = fptosi float [[TMP5]] to i32
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD7]], [[CONV8]]
// CHECK2-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ3barssssE2Sc, align 4
// CHECK2-NEXT:    [[CONV10:%.*]] = fptosi float [[TMP6]] to i32
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CONV10]]
// CHECK2-NEXT:    [[TMP7:%.*]] = load float, float* @_ZZ3barssssE2Sd, align 4
// CHECK2-NEXT:    [[CONV12:%.*]] = fptosi float [[TMP7]] to i32
// CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[ADD11]], [[CONV12]]
// CHECK2-NEXT:    ret i32 [[ADD13]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]]) #[[ATTR1]] {
// 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 i16*, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GB_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SB_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GC_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SC_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[D_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GD_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SD_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], 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 i16* [[A]], i16** [[A_ADDR]], align 8
// CHECK2-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 8
// CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK2-NEXT:    store i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP1]], align 2
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load double, double* @Gb, align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP6]], double* [[CONV1]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[GB_CASTED]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load float, float* @_ZZ3barssssE2Sb, align 4
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP8]], float* [[CONV2]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i64, i64* [[SB_CASTED]], align 8
// CHECK2-NEXT:    [[TMP10:%.*]] = load double, double* @Gc, align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP10]], double* [[CONV3]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = load i64, i64* [[GC_CASTED]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP2]], align 2
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP12]], i16* [[CONV4]], align 2
// CHECK2-NEXT:    [[TMP13:%.*]] = load i64, i64* [[C_CASTED]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ3barssssE2Sc, align 4
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP14]], float* [[CONV5]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[SC_CASTED]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP3]], align 2
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP16]], i16* [[CONV6]], align 2
// CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[D_CASTED]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = load double, double* @Gd, align 8
// CHECK2-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP18]], double* [[CONV7]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = load i64, i64* [[GD_CASTED]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = load float, float* @_ZZ3barssssE2Sd, align 4
// CHECK2-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP20]], float* [[CONV8]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i64, i64* [[SD_CASTED]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = load double, double* @Ga, align 8
// CHECK2-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP22]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       land.lhs.true:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK2-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP23]] to i32
// CHECK2-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[CONV9]], 0
// CHECK2-NEXT:    br i1 [[CMP10]], label [[LAND_LHS_TRUE11:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       land.lhs.true11:
// CHECK2-NEXT:    [[TMP24:%.*]] = load float, float* @_ZZ3barssssE2Sa, align 4
// CHECK2-NEXT:    [[CONV12:%.*]] = fpext float [[TMP24]] to double
// CHECK2-NEXT:    [[CMP13:%.*]] = fcmp ogt double [[CONV12]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP13]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP26]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP28]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK2-NEXT:    store i64 [[TMP7]], i64* [[TMP31]], align 8
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i64*
// CHECK2-NEXT:    store i64 [[TMP7]], i64* [[TMP33]], align 8
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP34]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i64*
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP36]], align 8
// CHECK2-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i64*
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP39]], align 8
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
// CHECK2-NEXT:    store i64 [[TMP11]], i64* [[TMP41]], align 8
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK2-NEXT:    store i64 [[TMP11]], i64* [[TMP43]], align 8
// CHECK2-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK2-NEXT:    store i8* null, i8** [[TMP44]], align 8
// CHECK2-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK2-NEXT:    store i64 [[TMP13]], i64* [[TMP46]], align 8
// CHECK2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK2-NEXT:    store i64 [[TMP13]], i64* [[TMP48]], align 8
// CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK2-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK2-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
// CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP51]], align 8
// CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK2-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
// CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP53]], align 8
// CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
// CHECK2-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK2-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK2-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i64*
// CHECK2-NEXT:    store i64 [[TMP17]], i64* [[TMP56]], align 8
// CHECK2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK2-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i64*
// CHECK2-NEXT:    store i64 [[TMP17]], i64* [[TMP58]], align 8
// CHECK2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 6
// CHECK2-NEXT:    store i8* null, i8** [[TMP59]], align 8
// CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK2-NEXT:    [[TMP61:%.*]] = bitcast i8** [[TMP60]] to i64*
// CHECK2-NEXT:    store i64 [[TMP19]], i64* [[TMP61]], align 8
// CHECK2-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK2-NEXT:    [[TMP63:%.*]] = bitcast i8** [[TMP62]] to i64*
// CHECK2-NEXT:    store i64 [[TMP19]], i64* [[TMP63]], align 8
// CHECK2-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 7
// CHECK2-NEXT:    store i8* null, i8** [[TMP64]], align 8
// CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK2-NEXT:    [[TMP66:%.*]] = bitcast i8** [[TMP65]] to i64*
// CHECK2-NEXT:    store i64 [[TMP21]], i64* [[TMP66]], align 8
// CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK2-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to i64*
// CHECK2-NEXT:    store i64 [[TMP21]], i64* [[TMP68]], align 8
// CHECK2-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 8
// CHECK2-NEXT:    store i8* null, i8** [[TMP69]], align 8
// CHECK2-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP72:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94.region_id, i32 9, i8** [[TMP70]], i8** [[TMP71]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes.3, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null)
// CHECK2-NEXT:    [[TMP73:%.*]] = icmp ne i32 [[TMP72]], 0
// CHECK2-NEXT:    br i1 [[TMP73]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94(i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94(i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94
// CHECK2-SAME: (i64 [[B:%.*]], i64 [[GB:%.*]], i64 [[SB:%.*]], i64 [[GC:%.*]], i64 [[C:%.*]], i64 [[SC:%.*]], i64 [[D:%.*]], i64 [[GD:%.*]], i64 [[SD:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GB_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SB_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GC_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SC_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GD_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SD_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GB]], i64* [[GB_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SB]], i64* [[SB_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GC]], i64* [[GC_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[C]], i64* [[C_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SC]], i64* [[SC_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[D]], i64* [[D_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GD]], i64* [[GD_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SD]], i64* [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_ADDR]] to double*
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_ADDR]] to float*
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_ADDR]] to double*
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_ADDR]] to float*
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_ADDR]] to double*
// CHECK2-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_ADDR]] to float*
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP1]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK2-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK2-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load double, double* [[CONV1]], align 8
// CHECK2-NEXT:    [[ADD11:%.*]] = fadd double [[TMP2]], 1.000000e+00
// CHECK2-NEXT:    store double [[ADD11]], double* [[CONV1]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load float, float* [[CONV2]], align 8
// CHECK2-NEXT:    [[CONV12:%.*]] = fpext float [[TMP3]] to double
// CHECK2-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK2-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK2-NEXT:    store float [[CONV14]], float* [[CONV2]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load double, double* [[CONV3]], align 8
// CHECK2-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP4]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       land.lhs.true:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV4]], align 8
// CHECK2-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP5]] to i32
// CHECK2-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK2-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       land.lhs.true17:
// CHECK2-NEXT:    [[TMP6:%.*]] = load float, float* [[CONV5]], align 8
// CHECK2-NEXT:    [[CONV18:%.*]] = fpext float [[TMP6]] to double
// CHECK2-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]])
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK2-NEXT:    call void @.omp_outlined..2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]]) #[[ATTR2]]
// CHECK2-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 8 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 8
// CHECK2-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK2-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK2-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK2-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK2-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK2-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK2-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK2-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK2-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z5tbar2ssss
// CHECK2-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_Z4tbarIsEiT_S0_S0_S0_(i16 signext [[TMP0]], i16 signext [[TMP1]], i16 signext [[TMP2]], i16 signext [[TMP3]])
// CHECK2-NEXT:    ret i32 [[CALL]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z4tbarIsEiT_S0_S0_S0_
// CHECK2-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] comdat {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK2-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, i16*, i16*, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i16* [[A_ADDR]], i16* [[B_ADDR]], i16* [[C_ADDR]], i16* [[D_ADDR]])
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK2-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV1]]
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CONV2]]
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK2-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP3]] to i32
// CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CONV4]]
// CHECK2-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sa, align 4
// CHECK2-NEXT:    [[CONV6:%.*]] = fptosi float [[TMP4]] to i32
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CONV6]]
// CHECK2-NEXT:    [[TMP5:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sb, align 4
// CHECK2-NEXT:    [[CONV8:%.*]] = fptosi float [[TMP5]] to i32
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD7]], [[CONV8]]
// CHECK2-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sc, align 4
// CHECK2-NEXT:    [[CONV10:%.*]] = fptosi float [[TMP6]] to i32
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CONV10]]
// CHECK2-NEXT:    [[TMP7:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sd, align 4
// CHECK2-NEXT:    [[CONV12:%.*]] = fptosi float [[TMP7]] to i32
// CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[ADD11]], [[CONV12]]
// CHECK2-NEXT:    ret i32 [[ADD13]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]]) #[[ATTR1]] {
// 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 i16*, align 8
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GB_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SB_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GC_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SC_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[D_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GD_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SD_CASTED:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 8
// CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], 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 i16* [[A]], i16** [[A_ADDR]], align 8
// CHECK2-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 8
// CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
// CHECK2-NEXT:    store i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[A_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i16*, i16** [[C_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP1]], align 2
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load double, double* @Gb, align 8
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP6]], double* [[CONV1]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[GB_CASTED]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sb, align 4
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP8]], float* [[CONV2]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i64, i64* [[SB_CASTED]], align 8
// CHECK2-NEXT:    [[TMP10:%.*]] = load double, double* @Gc, align 8
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP10]], double* [[CONV3]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = load i64, i64* [[GC_CASTED]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP2]], align 2
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP12]], i16* [[CONV4]], align 2
// CHECK2-NEXT:    [[TMP13:%.*]] = load i64, i64* [[C_CASTED]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sc, align 4
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP14]], float* [[CONV5]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[SC_CASTED]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP3]], align 2
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_CASTED]] to i16*
// CHECK2-NEXT:    store i16 [[TMP16]], i16* [[CONV6]], align 2
// CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[D_CASTED]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = load double, double* @Gd, align 8
// CHECK2-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_CASTED]] to double*
// CHECK2-NEXT:    store double [[TMP18]], double* [[CONV7]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = load i64, i64* [[GD_CASTED]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sd, align 4
// CHECK2-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_CASTED]] to float*
// CHECK2-NEXT:    store float [[TMP20]], float* [[CONV8]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i64, i64* [[SD_CASTED]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = load double, double* @Ga, align 8
// CHECK2-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP22]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       land.lhs.true:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK2-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP23]] to i32
// CHECK2-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[CONV9]], 0
// CHECK2-NEXT:    br i1 [[CMP10]], label [[LAND_LHS_TRUE11:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       land.lhs.true11:
// CHECK2-NEXT:    [[TMP24:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sa, align 4
// CHECK2-NEXT:    [[CONV12:%.*]] = fpext float [[TMP24]] to double
// CHECK2-NEXT:    [[CMP13:%.*]] = fcmp ogt double [[CONV12]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP13]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP26]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
// CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP28]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK2-NEXT:    store i8* null, i8** [[TMP29]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
// CHECK2-NEXT:    store i64 [[TMP7]], i64* [[TMP31]], align 8
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i64*
// CHECK2-NEXT:    store i64 [[TMP7]], i64* [[TMP33]], align 8
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
// CHECK2-NEXT:    store i8* null, i8** [[TMP34]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i64*
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP36]], align 8
// CHECK2-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i64*
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
// CHECK2-NEXT:    store i8* null, i8** [[TMP39]], align 8
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
// CHECK2-NEXT:    store i64 [[TMP11]], i64* [[TMP41]], align 8
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK2-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
// CHECK2-NEXT:    store i64 [[TMP11]], i64* [[TMP43]], align 8
// CHECK2-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
// CHECK2-NEXT:    store i8* null, i8** [[TMP44]], align 8
// CHECK2-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
// CHECK2-NEXT:    store i64 [[TMP13]], i64* [[TMP46]], align 8
// CHECK2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK2-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
// CHECK2-NEXT:    store i64 [[TMP13]], i64* [[TMP48]], align 8
// CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
// CHECK2-NEXT:    store i8* null, i8** [[TMP49]], align 8
// CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK2-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
// CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP51]], align 8
// CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK2-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
// CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP53]], align 8
// CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
// CHECK2-NEXT:    store i8* null, i8** [[TMP54]], align 8
// CHECK2-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK2-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i64*
// CHECK2-NEXT:    store i64 [[TMP17]], i64* [[TMP56]], align 8
// CHECK2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK2-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i64*
// CHECK2-NEXT:    store i64 [[TMP17]], i64* [[TMP58]], align 8
// CHECK2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 6
// CHECK2-NEXT:    store i8* null, i8** [[TMP59]], align 8
// CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK2-NEXT:    [[TMP61:%.*]] = bitcast i8** [[TMP60]] to i64*
// CHECK2-NEXT:    store i64 [[TMP19]], i64* [[TMP61]], align 8
// CHECK2-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK2-NEXT:    [[TMP63:%.*]] = bitcast i8** [[TMP62]] to i64*
// CHECK2-NEXT:    store i64 [[TMP19]], i64* [[TMP63]], align 8
// CHECK2-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 7
// CHECK2-NEXT:    store i8* null, i8** [[TMP64]], align 8
// CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK2-NEXT:    [[TMP66:%.*]] = bitcast i8** [[TMP65]] to i64*
// CHECK2-NEXT:    store i64 [[TMP21]], i64* [[TMP66]], align 8
// CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK2-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to i64*
// CHECK2-NEXT:    store i64 [[TMP21]], i64* [[TMP68]], align 8
// CHECK2-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 8
// CHECK2-NEXT:    store i8* null, i8** [[TMP69]], align 8
// CHECK2-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP72:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145.region_id, i32 9, i8** [[TMP70]], i8** [[TMP71]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null)
// CHECK2-NEXT:    [[TMP73:%.*]] = icmp ne i32 [[TMP72]], 0
// CHECK2-NEXT:    br i1 [[TMP73]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK2:       omp_offload.failed:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145(i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK2:       omp_offload.cont:
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145(i64 [[TMP5]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145
// CHECK2-SAME: (i64 [[B:%.*]], i64 [[GB:%.*]], i64 [[SB:%.*]], i64 [[GC:%.*]], i64 [[C:%.*]], i64 [[SC:%.*]], i64 [[D:%.*]], i64 [[GD:%.*]], i64 [[SD:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GB_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SB_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GC_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SC_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[GD_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[SD_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GB]], i64* [[GB_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SB]], i64* [[SB_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GC]], i64* [[GC_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[C]], i64* [[C_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SC]], i64* [[SC_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[D]], i64* [[D_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[GD]], i64* [[GD_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[SD]], i64* [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[GB_ADDR]] to double*
// CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[SB_ADDR]] to float*
// CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[GC_ADDR]] to double*
// CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[C_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[SC_ADDR]] to float*
// CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[D_ADDR]] to i16*
// CHECK2-NEXT:    [[CONV7:%.*]] = bitcast i64* [[GD_ADDR]] to double*
// CHECK2-NEXT:    [[CONV8:%.*]] = bitcast i64* [[SD_ADDR]] to float*
// CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
// CHECK2-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP1]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK2-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK2-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load double, double* [[CONV1]], align 8
// CHECK2-NEXT:    [[ADD11:%.*]] = fadd double [[TMP2]], 1.000000e+00
// CHECK2-NEXT:    store double [[ADD11]], double* [[CONV1]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load float, float* [[CONV2]], align 8
// CHECK2-NEXT:    [[CONV12:%.*]] = fpext float [[TMP3]] to double
// CHECK2-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK2-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK2-NEXT:    store float [[CONV14]], float* [[CONV2]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load double, double* [[CONV3]], align 8
// CHECK2-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP4]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK2:       land.lhs.true:
// CHECK2-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV4]], align 8
// CHECK2-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP5]] to i32
// CHECK2-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK2-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       land.lhs.true17:
// CHECK2-NEXT:    [[TMP6:%.*]] = load float, float* [[CONV5]], align 8
// CHECK2-NEXT:    [[CONV18:%.*]] = fpext float [[TMP6]] to double
// CHECK2-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK2:       omp_if.then:
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]])
// CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK2:       omp_if.else:
// CHECK2-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK2-NEXT:    call void @.omp_outlined..6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV6]], double* [[CONV7]], float* [[CONV8]]) #[[ATTR2]]
// CHECK2-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK2-NEXT:    br label [[OMP_IF_END]]
// CHECK2:       omp_if.end:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 8 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 8
// CHECK2-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 8
// CHECK2-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 8
// CHECK2-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK2-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK2-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK2-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK2-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK2-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK2-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK2-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK2-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK2-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK2-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z3foossss
// CHECK3-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SB_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[C_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SC_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[D_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SD_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP0]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load float, float* @_ZZ3foossssE2Sb, align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP2]], float* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[SB_CASTED]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP4]], i16* [[CONV2]], align 2
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[C_CASTED]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ3foossssE2Sc, align 4
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP6]], float* [[CONV3]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[SC_CASTED]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP8]], i16* [[CONV4]], align 2
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[D_CASTED]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load float, float* @_ZZ3foossssE2Sd, align 4
// CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP10]], float* [[CONV5]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[SD_CASTED]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = load double, double* @Ga, align 8
// CHECK3-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP12]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       land.lhs.true:
// CHECK3-NEXT:    [[TMP13:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP13]] to i32
// CHECK3-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[CONV6]], 0
// CHECK3-NEXT:    br i1 [[CMP7]], label [[LAND_LHS_TRUE8:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       land.lhs.true8:
// CHECK3-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ3foossssE2Sa, align 4
// CHECK3-NEXT:    [[CONV9:%.*]] = fpext float [[TMP14]] to double
// CHECK3-NEXT:    [[CMP10:%.*]] = fcmp ogt double [[CONV9]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP10]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP16]], align 4
// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
// CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP18]], align 4
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP19]], align 4
// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to double**
// CHECK3-NEXT:    store double* @Gb, double** [[TMP21]], align 4
// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to double**
// CHECK3-NEXT:    store double* @Gb, double** [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP24]], align 4
// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP26]], align 4
// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
// CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP28]], align 4
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP29]], align 4
// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to double**
// CHECK3-NEXT:    store double* @Gc, double** [[TMP31]], align 4
// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to double**
// CHECK3-NEXT:    store double* @Gc, double** [[TMP33]], align 4
// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP34]], align 4
// CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP36]], align 4
// CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP38]], align 4
// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK3-NEXT:    store i8* null, i8** [[TMP39]], align 4
// CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP41]], align 4
// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP43]], align 4
// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
// CHECK3-NEXT:    store i8* null, i8** [[TMP44]], align 4
// CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[TMP46]], align 4
// CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[TMP48]], align 4
// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 6
// CHECK3-NEXT:    store i8* null, i8** [[TMP49]], align 4
// CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to double**
// CHECK3-NEXT:    store double* @Gd, double** [[TMP51]], align 4
// CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to double**
// CHECK3-NEXT:    store double* @Gd, double** [[TMP53]], align 4
// CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 7
// CHECK3-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK3-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i32*
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[TMP56]], align 4
// CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i32*
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[TMP58]], align 4
// CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 8
// CHECK3-NEXT:    store i8* null, i8** [[TMP59]], align 4
// CHECK3-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP62:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49.region_id, i32 9, i8** [[TMP60]], i8** [[TMP61]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null)
// CHECK3-NEXT:    [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
// CHECK3-NEXT:    br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49(i32 [[TMP1]], double* @Gb, i32 [[TMP3]], double* @Gc, i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP9]], double* @Gd, i32 [[TMP11]]) #[[ATTR2:[0-9]+]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49(i32 [[TMP1]], double* @Gb, i32 [[TMP3]], double* @Gc, i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP9]], double* @Gd, i32 [[TMP11]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    [[TMP64:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    [[CONV11:%.*]] = sext i16 [[TMP64]] to i32
// CHECK3-NEXT:    [[TMP65:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP65]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV11]], [[CONV12]]
// CHECK3-NEXT:    [[TMP66:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    [[CONV13:%.*]] = sext i16 [[TMP66]] to i32
// CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 [[ADD]], [[CONV13]]
// CHECK3-NEXT:    [[TMP67:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP67]] to i32
// CHECK3-NEXT:    [[ADD16:%.*]] = add nsw i32 [[ADD14]], [[CONV15]]
// CHECK3-NEXT:    [[TMP68:%.*]] = load float, float* @_ZZ3foossssE2Sa, align 4
// CHECK3-NEXT:    [[CONV17:%.*]] = fptosi float [[TMP68]] to i32
// CHECK3-NEXT:    [[ADD18:%.*]] = add nsw i32 [[ADD16]], [[CONV17]]
// CHECK3-NEXT:    [[TMP69:%.*]] = load float, float* @_ZZ3foossssE2Sb, align 4
// CHECK3-NEXT:    [[CONV19:%.*]] = fptosi float [[TMP69]] to i32
// CHECK3-NEXT:    [[ADD20:%.*]] = add nsw i32 [[ADD18]], [[CONV19]]
// CHECK3-NEXT:    [[TMP70:%.*]] = load float, float* @_ZZ3foossssE2Sc, align 4
// CHECK3-NEXT:    [[CONV21:%.*]] = fptosi float [[TMP70]] to i32
// CHECK3-NEXT:    [[ADD22:%.*]] = add nsw i32 [[ADD20]], [[CONV21]]
// CHECK3-NEXT:    [[TMP71:%.*]] = load float, float* @_ZZ3foossssE2Sd, align 4
// CHECK3-NEXT:    [[CONV23:%.*]] = fptosi float [[TMP71]] to i32
// CHECK3-NEXT:    [[ADD24:%.*]] = add nsw i32 [[ADD22]], [[CONV23]]
// CHECK3-NEXT:    ret i32 [[ADD24]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49
// CHECK3-SAME: (i32 [[B:%.*]], double* nonnull align 4 dereferenceable(8) [[GB:%.*]], i32 [[SB:%.*]], double* nonnull align 4 dereferenceable(8) [[GC:%.*]], i32 [[C:%.*]], i32 [[SC:%.*]], i32 [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], i32 [[SD:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GB_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SB_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GC_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SC_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SD_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GB6:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[GC7:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[GD8:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GB]], double** [[GB_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SB]], i32* [[SB_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GC]], double** [[GC_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[C]], i32* [[C_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SC]], i32* [[SC_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[D]], i32* [[D_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SD]], i32* [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** [[GB_ADDR]], align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_ADDR]] to float*
// CHECK3-NEXT:    [[TMP2:%.*]] = load double*, double** [[GC_ADDR]], align 4
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_ADDR]] to i16*
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_ADDR]] to float*
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP3:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_ADDR]] to float*
// CHECK3-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK3-NEXT:    store double [[TMP4]], double* [[GB6]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP2]], align 8
// CHECK3-NEXT:    store double [[TMP5]], double* [[GC7]], align 8
// CHECK3-NEXT:    [[TMP6:%.*]] = load double, double* [[TMP3]], align 8
// CHECK3-NEXT:    store double [[TMP6]], double* [[GD8]], align 8
// CHECK3-NEXT:    [[TMP7:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP7]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK3-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK3-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load double, double* [[GB6]], align 8
// CHECK3-NEXT:    [[ADD11:%.*]] = fadd double [[TMP8]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD11]], double* [[GB6]], align 8
// CHECK3-NEXT:    [[TMP9:%.*]] = load float, float* [[CONV1]], align 4
// CHECK3-NEXT:    [[CONV12:%.*]] = fpext float [[TMP9]] to double
// CHECK3-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK3-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK3-NEXT:    store float [[CONV14]], float* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load double, double* [[GC7]], align 8
// CHECK3-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP10]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       land.lhs.true:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i16, i16* [[CONV2]], align 4
// CHECK3-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP11]] to i32
// CHECK3-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK3-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       land.lhs.true17:
// CHECK3-NEXT:    [[TMP12:%.*]] = load float, float* [[CONV3]], align 4
// CHECK3-NEXT:    [[CONV18:%.*]] = fpext float [[TMP12]] to double
// CHECK3-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined. to void (i32*, i32*, ...)*), i16* [[CONV4]], double* [[GD8]], float* [[CONV5]])
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK3-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV4]], double* [[GD8]], float* [[CONV5]]) #[[ATTR2]]
// CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK3-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK3-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK3-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK3-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK3-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK3-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK3-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK3-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z3barssss
// CHECK3-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, i16*, i16*, i16*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i16* [[A_ADDR]], i16* [[B_ADDR]], i16* [[C_ADDR]], i16* [[D_ADDR]])
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
// CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV1]]
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CONV2]]
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP3]] to i32
// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CONV4]]
// CHECK3-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ3barssssE2Sa, align 4
// CHECK3-NEXT:    [[CONV6:%.*]] = fptosi float [[TMP4]] to i32
// CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CONV6]]
// CHECK3-NEXT:    [[TMP5:%.*]] = load float, float* @_ZZ3barssssE2Sb, align 4
// CHECK3-NEXT:    [[CONV8:%.*]] = fptosi float [[TMP5]] to i32
// CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD7]], [[CONV8]]
// CHECK3-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ3barssssE2Sc, align 4
// CHECK3-NEXT:    [[CONV10:%.*]] = fptosi float [[TMP6]] to i32
// CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CONV10]]
// CHECK3-NEXT:    [[TMP7:%.*]] = load float, float* @_ZZ3barssssE2Sd, align 4
// CHECK3-NEXT:    [[CONV12:%.*]] = fptosi float [[TMP7]] to i32
// CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i32 [[ADD11]], [[CONV12]]
// CHECK3-NEXT:    ret i32 [[ADD13]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]]) #[[ATTR1]] {
// 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 i16*, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SB_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[C_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SC_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[D_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SD_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], 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 i16* [[A]], i16** [[A_ADDR]], align 4
// CHECK3-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
// CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK3-NEXT:    store i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP1]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ3barssssE2Sb, align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP6]], float* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[SB_CASTED]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i16, i16* [[TMP2]], align 2
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP8]], i16* [[CONV2]], align 2
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[C_CASTED]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load float, float* @_ZZ3barssssE2Sc, align 4
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP10]], float* [[CONV3]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[SC_CASTED]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP3]], align 2
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP12]], i16* [[CONV4]], align 2
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[D_CASTED]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ3barssssE2Sd, align 4
// CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP14]], float* [[CONV5]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[SD_CASTED]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = load double, double* @Ga, align 8
// CHECK3-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP16]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       land.lhs.true:
// CHECK3-NEXT:    [[TMP17:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK3-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP17]] to i32
// CHECK3-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[CONV6]], 0
// CHECK3-NEXT:    br i1 [[CMP7]], label [[LAND_LHS_TRUE8:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       land.lhs.true8:
// CHECK3-NEXT:    [[TMP18:%.*]] = load float, float* @_ZZ3barssssE2Sa, align 4
// CHECK3-NEXT:    [[CONV9:%.*]] = fpext float [[TMP18]] to double
// CHECK3-NEXT:    [[CMP10:%.*]] = fcmp ogt double [[CONV9]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP10]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to double**
// CHECK3-NEXT:    store double* @Gb, double** [[TMP25]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to double**
// CHECK3-NEXT:    store double* @Gb, double** [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP30]], align 4
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP32]], align 4
// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to double**
// CHECK3-NEXT:    store double* @Gc, double** [[TMP35]], align 4
// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to double**
// CHECK3-NEXT:    store double* @Gc, double** [[TMP37]], align 4
// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP38]], align 4
// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[TMP40]], align 4
// CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i32*
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[TMP42]], align 4
// CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK3-NEXT:    store i8* null, i8** [[TMP43]], align 4
// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i32*
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[TMP45]], align 4
// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[TMP47]], align 4
// CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
// CHECK3-NEXT:    store i8* null, i8** [[TMP48]], align 4
// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
// CHECK3-NEXT:    store i32 [[TMP13]], i32* [[TMP50]], align 4
// CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK3-NEXT:    store i32 [[TMP13]], i32* [[TMP52]], align 4
// CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 6
// CHECK3-NEXT:    store i8* null, i8** [[TMP53]], align 4
// CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to double**
// CHECK3-NEXT:    store double* @Gd, double** [[TMP55]], align 4
// CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to double**
// CHECK3-NEXT:    store double* @Gd, double** [[TMP57]], align 4
// CHECK3-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 7
// CHECK3-NEXT:    store i8* null, i8** [[TMP58]], align 4
// CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
// CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP60]], align 4
// CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP62]], align 4
// CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 8
// CHECK3-NEXT:    store i8* null, i8** [[TMP63]], align 4
// CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94.region_id, i32 9, i8** [[TMP64]], i8** [[TMP65]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes.3, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null)
// CHECK3-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
// CHECK3-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94(i32 [[TMP5]], double* @Gb, i32 [[TMP7]], double* @Gc, i32 [[TMP9]], i32 [[TMP11]], i32 [[TMP13]], double* @Gd, i32 [[TMP15]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94(i32 [[TMP5]], double* @Gb, i32 [[TMP7]], double* @Gc, i32 [[TMP9]], i32 [[TMP11]], i32 [[TMP13]], double* @Gd, i32 [[TMP15]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94
// CHECK3-SAME: (i32 [[B:%.*]], double* nonnull align 4 dereferenceable(8) [[GB:%.*]], i32 [[SB:%.*]], double* nonnull align 4 dereferenceable(8) [[GC:%.*]], i32 [[C:%.*]], i32 [[SC:%.*]], i32 [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], i32 [[SD:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GB_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SB_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GC_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SC_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SD_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GB6:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[GC7:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[GD8:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GB]], double** [[GB_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SB]], i32* [[SB_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GC]], double** [[GC_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[C]], i32* [[C_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SC]], i32* [[SC_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[D]], i32* [[D_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SD]], i32* [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** [[GB_ADDR]], align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_ADDR]] to float*
// CHECK3-NEXT:    [[TMP2:%.*]] = load double*, double** [[GC_ADDR]], align 4
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_ADDR]] to i16*
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_ADDR]] to float*
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP3:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_ADDR]] to float*
// CHECK3-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK3-NEXT:    store double [[TMP4]], double* [[GB6]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP2]], align 8
// CHECK3-NEXT:    store double [[TMP5]], double* [[GC7]], align 8
// CHECK3-NEXT:    [[TMP6:%.*]] = load double, double* [[TMP3]], align 8
// CHECK3-NEXT:    store double [[TMP6]], double* [[GD8]], align 8
// CHECK3-NEXT:    [[TMP7:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP7]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK3-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK3-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load double, double* [[GB6]], align 8
// CHECK3-NEXT:    [[ADD11:%.*]] = fadd double [[TMP8]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD11]], double* [[GB6]], align 8
// CHECK3-NEXT:    [[TMP9:%.*]] = load float, float* [[CONV1]], align 4
// CHECK3-NEXT:    [[CONV12:%.*]] = fpext float [[TMP9]] to double
// CHECK3-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK3-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK3-NEXT:    store float [[CONV14]], float* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load double, double* [[GC7]], align 8
// CHECK3-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP10]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       land.lhs.true:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i16, i16* [[CONV2]], align 4
// CHECK3-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP11]] to i32
// CHECK3-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK3-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       land.lhs.true17:
// CHECK3-NEXT:    [[TMP12:%.*]] = load float, float* [[CONV3]], align 4
// CHECK3-NEXT:    [[CONV18:%.*]] = fpext float [[TMP12]] to double
// CHECK3-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i16* [[CONV4]], double* [[GD8]], float* [[CONV5]])
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK3-NEXT:    call void @.omp_outlined..2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV4]], double* [[GD8]], float* [[CONV5]]) #[[ATTR2]]
// CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK3-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK3-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK3-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK3-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK3-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK3-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK3-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK3-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z5tbar2ssss
// CHECK3-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    [[CALL:%.*]] = call i32 @_Z4tbarIsEiT_S0_S0_S0_(i16 signext [[TMP0]], i16 signext [[TMP1]], i16 signext [[TMP2]], i16 signext [[TMP3]])
// CHECK3-NEXT:    ret i32 [[CALL]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z4tbarIsEiT_S0_S0_S0_
// CHECK3-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] comdat {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK3-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, i16*, i16*, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i16* [[A_ADDR]], i16* [[B_ADDR]], i16* [[C_ADDR]], i16* [[D_ADDR]])
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
// CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV1]]
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK3-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CONV2]]
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK3-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP3]] to i32
// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CONV4]]
// CHECK3-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sa, align 4
// CHECK3-NEXT:    [[CONV6:%.*]] = fptosi float [[TMP4]] to i32
// CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CONV6]]
// CHECK3-NEXT:    [[TMP5:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sb, align 4
// CHECK3-NEXT:    [[CONV8:%.*]] = fptosi float [[TMP5]] to i32
// CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD7]], [[CONV8]]
// CHECK3-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sc, align 4
// CHECK3-NEXT:    [[CONV10:%.*]] = fptosi float [[TMP6]] to i32
// CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CONV10]]
// CHECK3-NEXT:    [[TMP7:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sd, align 4
// CHECK3-NEXT:    [[CONV12:%.*]] = fptosi float [[TMP7]] to i32
// CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i32 [[ADD11]], [[CONV12]]
// CHECK3-NEXT:    ret i32 [[ADD13]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]]) #[[ATTR1]] {
// 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 i16*, align 4
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SB_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[C_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SC_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[D_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SD_CASTED:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], 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 i16* [[A]], i16** [[A_ADDR]], align 4
// CHECK3-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
// CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK3-NEXT:    store i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[A_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP1]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sb, align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP6]], float* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[SB_CASTED]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i16, i16* [[TMP2]], align 2
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP8]], i16* [[CONV2]], align 2
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[C_CASTED]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sc, align 4
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP10]], float* [[CONV3]], align 4
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[SC_CASTED]], align 4
// CHECK3-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP3]], align 2
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_CASTED]] to i16*
// CHECK3-NEXT:    store i16 [[TMP12]], i16* [[CONV4]], align 2
// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[D_CASTED]], align 4
// CHECK3-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sd, align 4
// CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_CASTED]] to float*
// CHECK3-NEXT:    store float [[TMP14]], float* [[CONV5]], align 4
// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[SD_CASTED]], align 4
// CHECK3-NEXT:    [[TMP16:%.*]] = load double, double* @Ga, align 8
// CHECK3-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP16]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       land.lhs.true:
// CHECK3-NEXT:    [[TMP17:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK3-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP17]] to i32
// CHECK3-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[CONV6]], 0
// CHECK3-NEXT:    br i1 [[CMP7]], label [[LAND_LHS_TRUE8:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       land.lhs.true8:
// CHECK3-NEXT:    [[TMP18:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sa, align 4
// CHECK3-NEXT:    [[CONV9:%.*]] = fpext float [[TMP18]] to double
// CHECK3-NEXT:    [[CMP10:%.*]] = fcmp ogt double [[CONV9]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP10]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK3-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to double**
// CHECK3-NEXT:    store double* @Gb, double** [[TMP25]], align 4
// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to double**
// CHECK3-NEXT:    store double* @Gb, double** [[TMP27]], align 4
// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK3-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP30]], align 4
// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK3-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i32*
// CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP32]], align 4
// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK3-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to double**
// CHECK3-NEXT:    store double* @Gc, double** [[TMP35]], align 4
// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to double**
// CHECK3-NEXT:    store double* @Gc, double** [[TMP37]], align 4
// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK3-NEXT:    store i8* null, i8** [[TMP38]], align 4
// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[TMP40]], align 4
// CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK3-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i32*
// CHECK3-NEXT:    store i32 [[TMP9]], i32* [[TMP42]], align 4
// CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK3-NEXT:    store i8* null, i8** [[TMP43]], align 4
// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i32*
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[TMP45]], align 4
// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK3-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
// CHECK3-NEXT:    store i32 [[TMP11]], i32* [[TMP47]], align 4
// CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
// CHECK3-NEXT:    store i8* null, i8** [[TMP48]], align 4
// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
// CHECK3-NEXT:    store i32 [[TMP13]], i32* [[TMP50]], align 4
// CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK3-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK3-NEXT:    store i32 [[TMP13]], i32* [[TMP52]], align 4
// CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 6
// CHECK3-NEXT:    store i8* null, i8** [[TMP53]], align 4
// CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to double**
// CHECK3-NEXT:    store double* @Gd, double** [[TMP55]], align 4
// CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK3-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to double**
// CHECK3-NEXT:    store double* @Gd, double** [[TMP57]], align 4
// CHECK3-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 7
// CHECK3-NEXT:    store i8* null, i8** [[TMP58]], align 4
// CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
// CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP60]], align 4
// CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK3-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP62]], align 4
// CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 8
// CHECK3-NEXT:    store i8* null, i8** [[TMP63]], align 4
// CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145.region_id, i32 9, i8** [[TMP64]], i8** [[TMP65]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null)
// CHECK3-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
// CHECK3-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK3:       omp_offload.failed:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145(i32 [[TMP5]], double* @Gb, i32 [[TMP7]], double* @Gc, i32 [[TMP9]], i32 [[TMP11]], i32 [[TMP13]], double* @Gd, i32 [[TMP15]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK3:       omp_offload.cont:
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145(i32 [[TMP5]], double* @Gb, i32 [[TMP7]], double* @Gc, i32 [[TMP9]], i32 [[TMP11]], i32 [[TMP13]], double* @Gd, i32 [[TMP15]]) #[[ATTR2]]
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145
// CHECK3-SAME: (i32 [[B:%.*]], double* nonnull align 4 dereferenceable(8) [[GB:%.*]], i32 [[SB:%.*]], double* nonnull align 4 dereferenceable(8) [[GC:%.*]], i32 [[C:%.*]], i32 [[SC:%.*]], i32 [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], i32 [[SD:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GB_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SB_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GC_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[SC_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SD_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[GB6:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[GC7:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[GD8:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GB]], double** [[GB_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SB]], i32* [[SB_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GC]], double** [[GC_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[C]], i32* [[C_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SC]], i32* [[SC_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[D]], i32* [[D_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    store i32 [[SD]], i32* [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** [[GB_ADDR]], align 4
// CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_ADDR]] to float*
// CHECK3-NEXT:    [[TMP2:%.*]] = load double*, double** [[GC_ADDR]], align 4
// CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_ADDR]] to i16*
// CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_ADDR]] to float*
// CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_ADDR]] to i16*
// CHECK3-NEXT:    [[TMP3:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_ADDR]] to float*
// CHECK3-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK3-NEXT:    store double [[TMP4]], double* [[GB6]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP2]], align 8
// CHECK3-NEXT:    store double [[TMP5]], double* [[GC7]], align 8
// CHECK3-NEXT:    [[TMP6:%.*]] = load double, double* [[TMP3]], align 8
// CHECK3-NEXT:    store double [[TMP6]], double* [[GD8]], align 8
// CHECK3-NEXT:    [[TMP7:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK3-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP7]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK3-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK3-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load double, double* [[GB6]], align 8
// CHECK3-NEXT:    [[ADD11:%.*]] = fadd double [[TMP8]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD11]], double* [[GB6]], align 8
// CHECK3-NEXT:    [[TMP9:%.*]] = load float, float* [[CONV1]], align 4
// CHECK3-NEXT:    [[CONV12:%.*]] = fpext float [[TMP9]] to double
// CHECK3-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK3-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK3-NEXT:    store float [[CONV14]], float* [[CONV1]], align 4
// CHECK3-NEXT:    [[TMP10:%.*]] = load double, double* [[GC7]], align 8
// CHECK3-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP10]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK3:       land.lhs.true:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i16, i16* [[CONV2]], align 4
// CHECK3-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP11]] to i32
// CHECK3-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK3-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       land.lhs.true17:
// CHECK3-NEXT:    [[TMP12:%.*]] = load float, float* [[CONV3]], align 4
// CHECK3-NEXT:    [[CONV18:%.*]] = fpext float [[TMP12]] to double
// CHECK3-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK3-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK3:       omp_if.then:
// CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i16* [[CONV4]], double* [[GD8]], float* [[CONV5]])
// CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK3:       omp_if.else:
// CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK3-NEXT:    call void @.omp_outlined..6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV4]], double* [[GD8]], float* [[CONV5]]) #[[ATTR2]]
// CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK3-NEXT:    br label [[OMP_IF_END]]
// CHECK3:       omp_if.end:
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK3-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK3-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 4
// CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK3-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK3-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK3-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK3-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK3-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK3-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK3-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK3-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK3-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK3-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z3foossss
// CHECK4-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SB_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[C_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SC_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[D_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SD_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP0]], i16* [[CONV]], align 2
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load float, float* @_ZZ3foossssE2Sb, align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP2]], float* [[CONV1]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[SB_CASTED]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP4]], i16* [[CONV2]], align 2
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[C_CASTED]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ3foossssE2Sc, align 4
// CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP6]], float* [[CONV3]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[SC_CASTED]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP8]], i16* [[CONV4]], align 2
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[D_CASTED]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load float, float* @_ZZ3foossssE2Sd, align 4
// CHECK4-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP10]], float* [[CONV5]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[SD_CASTED]], align 4
// CHECK4-NEXT:    [[TMP12:%.*]] = load double, double* @Ga, align 8
// CHECK4-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP12]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       land.lhs.true:
// CHECK4-NEXT:    [[TMP13:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP13]] to i32
// CHECK4-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[CONV6]], 0
// CHECK4-NEXT:    br i1 [[CMP7]], label [[LAND_LHS_TRUE8:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       land.lhs.true8:
// CHECK4-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ3foossssE2Sa, align 4
// CHECK4-NEXT:    [[CONV9:%.*]] = fpext float [[TMP14]] to double
// CHECK4-NEXT:    [[CMP10:%.*]] = fcmp ogt double [[CONV9]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP10]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP16]], align 4
// CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
// CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP18]], align 4
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP19]], align 4
// CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to double**
// CHECK4-NEXT:    store double* @Gb, double** [[TMP21]], align 4
// CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to double**
// CHECK4-NEXT:    store double* @Gb, double** [[TMP23]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP24]], align 4
// CHECK4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP26]], align 4
// CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
// CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP28]], align 4
// CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP29]], align 4
// CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to double**
// CHECK4-NEXT:    store double* @Gc, double** [[TMP31]], align 4
// CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to double**
// CHECK4-NEXT:    store double* @Gc, double** [[TMP33]], align 4
// CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK4-NEXT:    store i8* null, i8** [[TMP34]], align 4
// CHECK4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP36]], align 4
// CHECK4-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP38]], align 4
// CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK4-NEXT:    store i8* null, i8** [[TMP39]], align 4
// CHECK4-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i32*
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[TMP41]], align 4
// CHECK4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i32*
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[TMP43]], align 4
// CHECK4-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
// CHECK4-NEXT:    store i8* null, i8** [[TMP44]], align 4
// CHECK4-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
// CHECK4-NEXT:    store i32 [[TMP9]], i32* [[TMP46]], align 4
// CHECK4-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
// CHECK4-NEXT:    store i32 [[TMP9]], i32* [[TMP48]], align 4
// CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 6
// CHECK4-NEXT:    store i8* null, i8** [[TMP49]], align 4
// CHECK4-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK4-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to double**
// CHECK4-NEXT:    store double* @Gd, double** [[TMP51]], align 4
// CHECK4-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK4-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to double**
// CHECK4-NEXT:    store double* @Gd, double** [[TMP53]], align 4
// CHECK4-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 7
// CHECK4-NEXT:    store i8* null, i8** [[TMP54]], align 4
// CHECK4-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK4-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i32*
// CHECK4-NEXT:    store i32 [[TMP11]], i32* [[TMP56]], align 4
// CHECK4-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK4-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i32*
// CHECK4-NEXT:    store i32 [[TMP11]], i32* [[TMP58]], align 4
// CHECK4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 8
// CHECK4-NEXT:    store i8* null, i8** [[TMP59]], align 4
// CHECK4-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP62:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49.region_id, i32 9, i8** [[TMP60]], i8** [[TMP61]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null)
// CHECK4-NEXT:    [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
// CHECK4-NEXT:    br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49(i32 [[TMP1]], double* @Gb, i32 [[TMP3]], double* @Gc, i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP9]], double* @Gd, i32 [[TMP11]]) #[[ATTR2:[0-9]+]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49(i32 [[TMP1]], double* @Gb, i32 [[TMP3]], double* @Gc, i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP9]], double* @Gd, i32 [[TMP11]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    [[TMP64:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    [[CONV11:%.*]] = sext i16 [[TMP64]] to i32
// CHECK4-NEXT:    [[TMP65:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP65]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV11]], [[CONV12]]
// CHECK4-NEXT:    [[TMP66:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    [[CONV13:%.*]] = sext i16 [[TMP66]] to i32
// CHECK4-NEXT:    [[ADD14:%.*]] = add nsw i32 [[ADD]], [[CONV13]]
// CHECK4-NEXT:    [[TMP67:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP67]] to i32
// CHECK4-NEXT:    [[ADD16:%.*]] = add nsw i32 [[ADD14]], [[CONV15]]
// CHECK4-NEXT:    [[TMP68:%.*]] = load float, float* @_ZZ3foossssE2Sa, align 4
// CHECK4-NEXT:    [[CONV17:%.*]] = fptosi float [[TMP68]] to i32
// CHECK4-NEXT:    [[ADD18:%.*]] = add nsw i32 [[ADD16]], [[CONV17]]
// CHECK4-NEXT:    [[TMP69:%.*]] = load float, float* @_ZZ3foossssE2Sb, align 4
// CHECK4-NEXT:    [[CONV19:%.*]] = fptosi float [[TMP69]] to i32
// CHECK4-NEXT:    [[ADD20:%.*]] = add nsw i32 [[ADD18]], [[CONV19]]
// CHECK4-NEXT:    [[TMP70:%.*]] = load float, float* @_ZZ3foossssE2Sc, align 4
// CHECK4-NEXT:    [[CONV21:%.*]] = fptosi float [[TMP70]] to i32
// CHECK4-NEXT:    [[ADD22:%.*]] = add nsw i32 [[ADD20]], [[CONV21]]
// CHECK4-NEXT:    [[TMP71:%.*]] = load float, float* @_ZZ3foossssE2Sd, align 4
// CHECK4-NEXT:    [[CONV23:%.*]] = fptosi float [[TMP71]] to i32
// CHECK4-NEXT:    [[ADD24:%.*]] = add nsw i32 [[ADD22]], [[CONV23]]
// CHECK4-NEXT:    ret i32 [[ADD24]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foossss_l49
// CHECK4-SAME: (i32 [[B:%.*]], double* nonnull align 4 dereferenceable(8) [[GB:%.*]], i32 [[SB:%.*]], double* nonnull align 4 dereferenceable(8) [[GC:%.*]], i32 [[C:%.*]], i32 [[SC:%.*]], i32 [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], i32 [[SD:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GB_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SB_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GC_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SC_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SD_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GB6:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[GC7:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[GD8:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GB]], double** [[GB_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SB]], i32* [[SB_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GC]], double** [[GC_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[C]], i32* [[C_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SC]], i32* [[SC_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[D]], i32* [[D_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SD]], i32* [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** [[GB_ADDR]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_ADDR]] to float*
// CHECK4-NEXT:    [[TMP2:%.*]] = load double*, double** [[GC_ADDR]], align 4
// CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_ADDR]] to i16*
// CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_ADDR]] to float*
// CHECK4-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP3:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_ADDR]] to float*
// CHECK4-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK4-NEXT:    store double [[TMP4]], double* [[GB6]], align 8
// CHECK4-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP2]], align 8
// CHECK4-NEXT:    store double [[TMP5]], double* [[GC7]], align 8
// CHECK4-NEXT:    [[TMP6:%.*]] = load double, double* [[TMP3]], align 8
// CHECK4-NEXT:    store double [[TMP6]], double* [[GD8]], align 8
// CHECK4-NEXT:    [[TMP7:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP7]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK4-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK4-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load double, double* [[GB6]], align 8
// CHECK4-NEXT:    [[ADD11:%.*]] = fadd double [[TMP8]], 1.000000e+00
// CHECK4-NEXT:    store double [[ADD11]], double* [[GB6]], align 8
// CHECK4-NEXT:    [[TMP9:%.*]] = load float, float* [[CONV1]], align 4
// CHECK4-NEXT:    [[CONV12:%.*]] = fpext float [[TMP9]] to double
// CHECK4-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK4-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK4-NEXT:    store float [[CONV14]], float* [[CONV1]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load double, double* [[GC7]], align 8
// CHECK4-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP10]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       land.lhs.true:
// CHECK4-NEXT:    [[TMP11:%.*]] = load i16, i16* [[CONV2]], align 4
// CHECK4-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP11]] to i32
// CHECK4-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK4-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       land.lhs.true17:
// CHECK4-NEXT:    [[TMP12:%.*]] = load float, float* [[CONV3]], align 4
// CHECK4-NEXT:    [[CONV18:%.*]] = fpext float [[TMP12]] to double
// CHECK4-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined. to void (i32*, i32*, ...)*), i16* [[CONV4]], double* [[GD8]], float* [[CONV5]])
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK4-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV4]], double* [[GD8]], float* [[CONV5]]) #[[ATTR2]]
// CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK4-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK4-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK4-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK4-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK4-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK4-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK4-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK4-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK4-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK4-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z3barssss
// CHECK4-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, i16*, i16*, i16*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i16* [[A_ADDR]], i16* [[B_ADDR]], i16* [[C_ADDR]], i16* [[D_ADDR]])
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
// CHECK4-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV1]]
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CONV2]]
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP3]] to i32
// CHECK4-NEXT:    [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CONV4]]
// CHECK4-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ3barssssE2Sa, align 4
// CHECK4-NEXT:    [[CONV6:%.*]] = fptosi float [[TMP4]] to i32
// CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CONV6]]
// CHECK4-NEXT:    [[TMP5:%.*]] = load float, float* @_ZZ3barssssE2Sb, align 4
// CHECK4-NEXT:    [[CONV8:%.*]] = fptosi float [[TMP5]] to i32
// CHECK4-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD7]], [[CONV8]]
// CHECK4-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ3barssssE2Sc, align 4
// CHECK4-NEXT:    [[CONV10:%.*]] = fptosi float [[TMP6]] to i32
// CHECK4-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CONV10]]
// CHECK4-NEXT:    [[TMP7:%.*]] = load float, float* @_ZZ3barssssE2Sd, align 4
// CHECK4-NEXT:    [[CONV12:%.*]] = fptosi float [[TMP7]] to i32
// CHECK4-NEXT:    [[ADD13:%.*]] = add nsw i32 [[ADD11]], [[CONV12]]
// CHECK4-NEXT:    ret i32 [[ADD13]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]]) #[[ATTR1]] {
// 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 i16*, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SB_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[C_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SC_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[D_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SD_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], 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 i16* [[A]], i16** [[A_ADDR]], align 4
// CHECK4-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
// CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK4-NEXT:    store i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP1]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ3barssssE2Sb, align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP6]], float* [[CONV1]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[SB_CASTED]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i16, i16* [[TMP2]], align 2
// CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP8]], i16* [[CONV2]], align 2
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[C_CASTED]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load float, float* @_ZZ3barssssE2Sc, align 4
// CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP10]], float* [[CONV3]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[SC_CASTED]], align 4
// CHECK4-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP3]], align 2
// CHECK4-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP12]], i16* [[CONV4]], align 2
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[D_CASTED]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ3barssssE2Sd, align 4
// CHECK4-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP14]], float* [[CONV5]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[SD_CASTED]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = load double, double* @Ga, align 8
// CHECK4-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP16]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       land.lhs.true:
// CHECK4-NEXT:    [[TMP17:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK4-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP17]] to i32
// CHECK4-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[CONV6]], 0
// CHECK4-NEXT:    br i1 [[CMP7]], label [[LAND_LHS_TRUE8:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       land.lhs.true8:
// CHECK4-NEXT:    [[TMP18:%.*]] = load float, float* @_ZZ3barssssE2Sa, align 4
// CHECK4-NEXT:    [[CONV9:%.*]] = fpext float [[TMP18]] to double
// CHECK4-NEXT:    [[CMP10:%.*]] = fcmp ogt double [[CONV9]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP10]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
// CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to double**
// CHECK4-NEXT:    store double* @Gb, double** [[TMP25]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to double**
// CHECK4-NEXT:    store double* @Gb, double** [[TMP27]], align 4
// CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[TMP30]], align 4
// CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i32*
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[TMP32]], align 4
// CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to double**
// CHECK4-NEXT:    store double* @Gc, double** [[TMP35]], align 4
// CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to double**
// CHECK4-NEXT:    store double* @Gc, double** [[TMP37]], align 4
// CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK4-NEXT:    store i8* null, i8** [[TMP38]], align 4
// CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
// CHECK4-NEXT:    store i32 [[TMP9]], i32* [[TMP40]], align 4
// CHECK4-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i32*
// CHECK4-NEXT:    store i32 [[TMP9]], i32* [[TMP42]], align 4
// CHECK4-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK4-NEXT:    store i8* null, i8** [[TMP43]], align 4
// CHECK4-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i32*
// CHECK4-NEXT:    store i32 [[TMP11]], i32* [[TMP45]], align 4
// CHECK4-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
// CHECK4-NEXT:    store i32 [[TMP11]], i32* [[TMP47]], align 4
// CHECK4-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
// CHECK4-NEXT:    store i8* null, i8** [[TMP48]], align 4
// CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
// CHECK4-NEXT:    store i32 [[TMP13]], i32* [[TMP50]], align 4
// CHECK4-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK4-NEXT:    store i32 [[TMP13]], i32* [[TMP52]], align 4
// CHECK4-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 6
// CHECK4-NEXT:    store i8* null, i8** [[TMP53]], align 4
// CHECK4-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK4-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to double**
// CHECK4-NEXT:    store double* @Gd, double** [[TMP55]], align 4
// CHECK4-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK4-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to double**
// CHECK4-NEXT:    store double* @Gd, double** [[TMP57]], align 4
// CHECK4-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 7
// CHECK4-NEXT:    store i8* null, i8** [[TMP58]], align 4
// CHECK4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK4-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
// CHECK4-NEXT:    store i32 [[TMP15]], i32* [[TMP60]], align 4
// CHECK4-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK4-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK4-NEXT:    store i32 [[TMP15]], i32* [[TMP62]], align 4
// CHECK4-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 8
// CHECK4-NEXT:    store i8* null, i8** [[TMP63]], align 4
// CHECK4-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94.region_id, i32 9, i8** [[TMP64]], i8** [[TMP65]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes.3, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null)
// CHECK4-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
// CHECK4-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94(i32 [[TMP5]], double* @Gb, i32 [[TMP7]], double* @Gc, i32 [[TMP9]], i32 [[TMP11]], i32 [[TMP13]], double* @Gd, i32 [[TMP15]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94(i32 [[TMP5]], double* @Gb, i32 [[TMP7]], double* @Gc, i32 [[TMP9]], i32 [[TMP11]], i32 [[TMP13]], double* @Gd, i32 [[TMP15]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3barssss_l94
// CHECK4-SAME: (i32 [[B:%.*]], double* nonnull align 4 dereferenceable(8) [[GB:%.*]], i32 [[SB:%.*]], double* nonnull align 4 dereferenceable(8) [[GC:%.*]], i32 [[C:%.*]], i32 [[SC:%.*]], i32 [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], i32 [[SD:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GB_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SB_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GC_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SC_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SD_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GB6:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[GC7:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[GD8:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GB]], double** [[GB_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SB]], i32* [[SB_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GC]], double** [[GC_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[C]], i32* [[C_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SC]], i32* [[SC_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[D]], i32* [[D_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SD]], i32* [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** [[GB_ADDR]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_ADDR]] to float*
// CHECK4-NEXT:    [[TMP2:%.*]] = load double*, double** [[GC_ADDR]], align 4
// CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_ADDR]] to i16*
// CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_ADDR]] to float*
// CHECK4-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP3:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_ADDR]] to float*
// CHECK4-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK4-NEXT:    store double [[TMP4]], double* [[GB6]], align 8
// CHECK4-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP2]], align 8
// CHECK4-NEXT:    store double [[TMP5]], double* [[GC7]], align 8
// CHECK4-NEXT:    [[TMP6:%.*]] = load double, double* [[TMP3]], align 8
// CHECK4-NEXT:    store double [[TMP6]], double* [[GD8]], align 8
// CHECK4-NEXT:    [[TMP7:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP7]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK4-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK4-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load double, double* [[GB6]], align 8
// CHECK4-NEXT:    [[ADD11:%.*]] = fadd double [[TMP8]], 1.000000e+00
// CHECK4-NEXT:    store double [[ADD11]], double* [[GB6]], align 8
// CHECK4-NEXT:    [[TMP9:%.*]] = load float, float* [[CONV1]], align 4
// CHECK4-NEXT:    [[CONV12:%.*]] = fpext float [[TMP9]] to double
// CHECK4-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK4-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK4-NEXT:    store float [[CONV14]], float* [[CONV1]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load double, double* [[GC7]], align 8
// CHECK4-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP10]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       land.lhs.true:
// CHECK4-NEXT:    [[TMP11:%.*]] = load i16, i16* [[CONV2]], align 4
// CHECK4-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP11]] to i32
// CHECK4-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK4-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       land.lhs.true17:
// CHECK4-NEXT:    [[TMP12:%.*]] = load float, float* [[CONV3]], align 4
// CHECK4-NEXT:    [[CONV18:%.*]] = fpext float [[TMP12]] to double
// CHECK4-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i16* [[CONV4]], double* [[GD8]], float* [[CONV5]])
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK4-NEXT:    call void @.omp_outlined..2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV4]], double* [[GD8]], float* [[CONV5]]) #[[ATTR2]]
// CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK4-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK4-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK4-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK4-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK4-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK4-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK4-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK4-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK4-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK4-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z5tbar2ssss
// CHECK4-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    [[CALL:%.*]] = call i32 @_Z4tbarIsEiT_S0_S0_S0_(i16 signext [[TMP0]], i16 signext [[TMP1]], i16 signext [[TMP2]], i16 signext [[TMP3]])
// CHECK4-NEXT:    ret i32 [[CALL]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z4tbarIsEiT_S0_S0_S0_
// CHECK4-SAME: (i16 signext [[A:%.*]], i16 signext [[B:%.*]], i16 signext [[C:%.*]], i16 signext [[D:%.*]]) #[[ATTR0]] comdat {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16, align 2
// CHECK4-NEXT:    store i16 [[A]], i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[B]], i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[C]], i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    store i16 [[D]], i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, i16*, i16*, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i16* [[A_ADDR]], i16* [[B_ADDR]], i16* [[C_ADDR]], i16* [[D_ADDR]])
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A_ADDR]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
// CHECK4-NEXT:    [[TMP1:%.*]] = load i16, i16* [[B_ADDR]], align 2
// CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV1]]
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[C_ADDR]], align 2
// CHECK4-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
// CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[ADD]], [[CONV2]]
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[D_ADDR]], align 2
// CHECK4-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP3]] to i32
// CHECK4-NEXT:    [[ADD5:%.*]] = add nsw i32 [[ADD3]], [[CONV4]]
// CHECK4-NEXT:    [[TMP4:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sa, align 4
// CHECK4-NEXT:    [[CONV6:%.*]] = fptosi float [[TMP4]] to i32
// CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[ADD5]], [[CONV6]]
// CHECK4-NEXT:    [[TMP5:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sb, align 4
// CHECK4-NEXT:    [[CONV8:%.*]] = fptosi float [[TMP5]] to i32
// CHECK4-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD7]], [[CONV8]]
// CHECK4-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sc, align 4
// CHECK4-NEXT:    [[CONV10:%.*]] = fptosi float [[TMP6]] to i32
// CHECK4-NEXT:    [[ADD11:%.*]] = add nsw i32 [[ADD9]], [[CONV10]]
// CHECK4-NEXT:    [[TMP7:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sd, align 4
// CHECK4-NEXT:    [[CONV12:%.*]] = fptosi float [[TMP7]] to i32
// CHECK4-NEXT:    [[ADD13:%.*]] = add nsw i32 [[ADD11]], [[CONV12]]
// CHECK4-NEXT:    ret i32 [[ADD13]]
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[A:%.*]], i16* nonnull align 2 dereferenceable(2) [[B:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]]) #[[ATTR1]] {
// 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 i16*, align 4
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SB_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[C_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SC_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[D_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SD_CASTED:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [9 x i8*], align 4
// CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [9 x i8*], 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 i16* [[A]], i16** [[A_ADDR]], align 4
// CHECK4-NEXT:    store i16* [[B]], i16** [[B_ADDR]], align 4
// CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
// CHECK4-NEXT:    store i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[A_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load i16*, i16** [[B_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load i16*, i16** [[C_ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP1]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sb, align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP6]], float* [[CONV1]], align 4
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[SB_CASTED]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i16, i16* [[TMP2]], align 2
// CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP8]], i16* [[CONV2]], align 2
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[C_CASTED]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sc, align 4
// CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP10]], float* [[CONV3]], align 4
// CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[SC_CASTED]], align 4
// CHECK4-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP3]], align 2
// CHECK4-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_CASTED]] to i16*
// CHECK4-NEXT:    store i16 [[TMP12]], i16* [[CONV4]], align 2
// CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[D_CASTED]], align 4
// CHECK4-NEXT:    [[TMP14:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sd, align 4
// CHECK4-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_CASTED]] to float*
// CHECK4-NEXT:    store float [[TMP14]], float* [[CONV5]], align 4
// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[SD_CASTED]], align 4
// CHECK4-NEXT:    [[TMP16:%.*]] = load double, double* @Ga, align 8
// CHECK4-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP16]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       land.lhs.true:
// CHECK4-NEXT:    [[TMP17:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK4-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP17]] to i32
// CHECK4-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[CONV6]], 0
// CHECK4-NEXT:    br i1 [[CMP7]], label [[LAND_LHS_TRUE8:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       land.lhs.true8:
// CHECK4-NEXT:    [[TMP18:%.*]] = load float, float* @_ZZ4tbarIsEiT_S0_S0_S0_E2Sa, align 4
// CHECK4-NEXT:    [[CONV9:%.*]] = fpext float [[TMP18]] to double
// CHECK4-NEXT:    [[CMP10:%.*]] = fcmp ogt double [[CONV9]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP10]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
// CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
// CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
// CHECK4-NEXT:    store i8* null, i8** [[TMP23]], align 4
// CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to double**
// CHECK4-NEXT:    store double* @Gb, double** [[TMP25]], align 4
// CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
// CHECK4-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to double**
// CHECK4-NEXT:    store double* @Gb, double** [[TMP27]], align 4
// CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
// CHECK4-NEXT:    store i8* null, i8** [[TMP28]], align 4
// CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[TMP30]], align 4
// CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
// CHECK4-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i32*
// CHECK4-NEXT:    store i32 [[TMP7]], i32* [[TMP32]], align 4
// CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
// CHECK4-NEXT:    store i8* null, i8** [[TMP33]], align 4
// CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to double**
// CHECK4-NEXT:    store double* @Gc, double** [[TMP35]], align 4
// CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to double**
// CHECK4-NEXT:    store double* @Gc, double** [[TMP37]], align 4
// CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
// CHECK4-NEXT:    store i8* null, i8** [[TMP38]], align 4
// CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
// CHECK4-NEXT:    store i32 [[TMP9]], i32* [[TMP40]], align 4
// CHECK4-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
// CHECK4-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i32*
// CHECK4-NEXT:    store i32 [[TMP9]], i32* [[TMP42]], align 4
// CHECK4-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
// CHECK4-NEXT:    store i8* null, i8** [[TMP43]], align 4
// CHECK4-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i32*
// CHECK4-NEXT:    store i32 [[TMP11]], i32* [[TMP45]], align 4
// CHECK4-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
// CHECK4-NEXT:    store i32 [[TMP11]], i32* [[TMP47]], align 4
// CHECK4-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
// CHECK4-NEXT:    store i8* null, i8** [[TMP48]], align 4
// CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
// CHECK4-NEXT:    store i32 [[TMP13]], i32* [[TMP50]], align 4
// CHECK4-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
// CHECK4-NEXT:    store i32 [[TMP13]], i32* [[TMP52]], align 4
// CHECK4-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 6
// CHECK4-NEXT:    store i8* null, i8** [[TMP53]], align 4
// CHECK4-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 7
// CHECK4-NEXT:    [[TMP55:%.*]] = bitcast i8** [[TMP54]] to double**
// CHECK4-NEXT:    store double* @Gd, double** [[TMP55]], align 4
// CHECK4-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 7
// CHECK4-NEXT:    [[TMP57:%.*]] = bitcast i8** [[TMP56]] to double**
// CHECK4-NEXT:    store double* @Gd, double** [[TMP57]], align 4
// CHECK4-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 7
// CHECK4-NEXT:    store i8* null, i8** [[TMP58]], align 4
// CHECK4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 8
// CHECK4-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
// CHECK4-NEXT:    store i32 [[TMP15]], i32* [[TMP60]], align 4
// CHECK4-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 8
// CHECK4-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
// CHECK4-NEXT:    store i32 [[TMP15]], i32* [[TMP62]], align 4
// CHECK4-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 8
// CHECK4-NEXT:    store i8* null, i8** [[TMP63]], align 4
// CHECK4-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK4-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145.region_id, i32 9, i8** [[TMP64]], i8** [[TMP65]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null)
// CHECK4-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
// CHECK4-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK4:       omp_offload.failed:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145(i32 [[TMP5]], double* @Gb, i32 [[TMP7]], double* @Gc, i32 [[TMP9]], i32 [[TMP11]], i32 [[TMP13]], double* @Gd, i32 [[TMP15]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK4:       omp_offload.cont:
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145(i32 [[TMP5]], double* @Gb, i32 [[TMP7]], double* @Gc, i32 [[TMP9]], i32 [[TMP11]], i32 [[TMP13]], double* @Gd, i32 [[TMP15]]) #[[ATTR2]]
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z4tbarIsEiT_S0_S0_S0__l145
// CHECK4-SAME: (i32 [[B:%.*]], double* nonnull align 4 dereferenceable(8) [[GB:%.*]], i32 [[SB:%.*]], double* nonnull align 4 dereferenceable(8) [[GC:%.*]], i32 [[C:%.*]], i32 [[SC:%.*]], i32 [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], i32 [[SD:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GB_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SB_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GC_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[SC_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SD_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[GB6:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[GC7:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[GD8:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
// CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GB]], double** [[GB_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SB]], i32* [[SB_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GC]], double** [[GC_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[C]], i32* [[C_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SC]], i32* [[SC_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[D]], i32* [[D_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    store i32 [[SD]], i32* [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** [[GB_ADDR]], align 4
// CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[SB_ADDR]] to float*
// CHECK4-NEXT:    [[TMP2:%.*]] = load double*, double** [[GC_ADDR]], align 4
// CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[C_ADDR]] to i16*
// CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[SC_ADDR]] to float*
// CHECK4-NEXT:    [[CONV4:%.*]] = bitcast i32* [[D_ADDR]] to i16*
// CHECK4-NEXT:    [[TMP3:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    [[CONV5:%.*]] = bitcast i32* [[SD_ADDR]] to float*
// CHECK4-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK4-NEXT:    store double [[TMP4]], double* [[GB6]], align 8
// CHECK4-NEXT:    [[TMP5:%.*]] = load double, double* [[TMP2]], align 8
// CHECK4-NEXT:    store double [[TMP5]], double* [[GC7]], align 8
// CHECK4-NEXT:    [[TMP6:%.*]] = load double, double* [[TMP3]], align 8
// CHECK4-NEXT:    store double [[TMP6]], double* [[GD8]], align 8
// CHECK4-NEXT:    [[TMP7:%.*]] = load i16, i16* [[CONV]], align 4
// CHECK4-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP7]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV9]], 1
// CHECK4-NEXT:    [[CONV10:%.*]] = trunc i32 [[ADD]] to i16
// CHECK4-NEXT:    store i16 [[CONV10]], i16* [[CONV]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load double, double* [[GB6]], align 8
// CHECK4-NEXT:    [[ADD11:%.*]] = fadd double [[TMP8]], 1.000000e+00
// CHECK4-NEXT:    store double [[ADD11]], double* [[GB6]], align 8
// CHECK4-NEXT:    [[TMP9:%.*]] = load float, float* [[CONV1]], align 4
// CHECK4-NEXT:    [[CONV12:%.*]] = fpext float [[TMP9]] to double
// CHECK4-NEXT:    [[ADD13:%.*]] = fadd double [[CONV12]], 1.000000e+00
// CHECK4-NEXT:    [[CONV14:%.*]] = fptrunc double [[ADD13]] to float
// CHECK4-NEXT:    store float [[CONV14]], float* [[CONV1]], align 4
// CHECK4-NEXT:    [[TMP10:%.*]] = load double, double* [[GC7]], align 8
// CHECK4-NEXT:    [[CMP:%.*]] = fcmp ogt double [[TMP10]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_IF_ELSE:%.*]]
// CHECK4:       land.lhs.true:
// CHECK4-NEXT:    [[TMP11:%.*]] = load i16, i16* [[CONV2]], align 4
// CHECK4-NEXT:    [[CONV15:%.*]] = sext i16 [[TMP11]] to i32
// CHECK4-NEXT:    [[CMP16:%.*]] = icmp sgt i32 [[CONV15]], 0
// CHECK4-NEXT:    br i1 [[CMP16]], label [[LAND_LHS_TRUE17:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       land.lhs.true17:
// CHECK4-NEXT:    [[TMP12:%.*]] = load float, float* [[CONV3]], align 4
// CHECK4-NEXT:    [[CONV18:%.*]] = fpext float [[TMP12]] to double
// CHECK4-NEXT:    [[CMP19:%.*]] = fcmp ogt double [[CONV18]], 0.000000e+00
// CHECK4-NEXT:    br i1 [[CMP19]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE]]
// CHECK4:       omp_if.then:
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i16*, double*, float*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i16* [[CONV4]], double* [[GD8]], float* [[CONV5]])
// CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
// CHECK4:       omp_if.else:
// CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
// CHECK4-NEXT:    call void @.omp_outlined..6(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i16* [[CONV4]], double* [[GD8]], float* [[CONV5]]) #[[ATTR2]]
// CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
// CHECK4-NEXT:    br label [[OMP_IF_END]]
// CHECK4:       omp_if.end:
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..6
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i16* nonnull align 2 dereferenceable(2) [[D:%.*]], double* nonnull align 4 dereferenceable(8) [[GD:%.*]], float* nonnull align 4 dereferenceable(4) [[SD:%.*]]) #[[ATTR1]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
// CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca i16*, align 4
// CHECK4-NEXT:    [[GD_ADDR:%.*]] = alloca double*, align 4
// CHECK4-NEXT:    [[SD_ADDR:%.*]] = alloca float*, 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 i16* [[D]], i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    store double* [[GD]], double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    store float* [[SD]], float** [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i16*, i16** [[D_ADDR]], align 4
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** [[GD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[SD_ADDR]], align 4
// CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[TMP0]], align 2
// CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
// CHECK4-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
// CHECK4-NEXT:    store i16 [[CONV1]], i16* [[TMP0]], align 2
// CHECK4-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP1]], align 8
// CHECK4-NEXT:    [[ADD2:%.*]] = fadd double [[TMP4]], 1.000000e+00
// CHECK4-NEXT:    store double [[ADD2]], double* [[TMP1]], align 8
// CHECK4-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP2]], align 4
// CHECK4-NEXT:    [[CONV3:%.*]] = fpext float [[TMP5]] to double
// CHECK4-NEXT:    [[ADD4:%.*]] = fadd double [[CONV3]], 1.000000e+00
// CHECK4-NEXT:    [[CONV5:%.*]] = fptrunc double [[ADD4]] to float
// CHECK4-NEXT:    store float [[CONV5]], float* [[TMP2]], align 4
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
// CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
// CHECK4-NEXT:    ret void
//