test/OpenMP/target_teams_distribute_parallel_for_simd_codegen.cpp - llvm-project/clang - Git at Google

 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
 // Test host codegen.
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3

 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5
 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK7
 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK7

 // Test target codegen - host bc file has to be created first. (no significant differences with host version of target region)
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK9
 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK11
 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11

 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK5
 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK7
 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK7

 #ifdef CK1

 int target_teams_fun(int *g){
   int n = 1000;
   int a[1000];
   int te = n / 128;
   int th = 128;
 // discard n_addr
 // discard capture expressions for te and th

   int i;
 #pragma omp target teams distribute parallel for simd num_teams(te), thread_limit(th) aligned(a : 8) safelen(16) simdlen(4) linear(i : n)
   for(i = 0; i < n; i++) {
     a[i] = 0;
   }

   {{{
   #pragma omp target teams distribute parallel for simd is_device_ptr(g) simdlen(8)
   for(int i = 0; i < n; i++) {
     a[i] = g[0];
   }
   }}}

   // outlined target regions


   return a[0];
 }

 void test_target_teams_atomic() {
   int x = 0;
 #pragma omp target teams distribute parallel for simd map(tofrom: x) shared(x)
   for (int i = 0; i < 1024; i++) {
 #pragma omp atomic update
     x += 1;
   }
 }


 #endif // CK1
 #endif // HEADER


 // CHECK1-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
 // CHECK1-SAME: (ptr noundef [[G:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[N:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
 // CHECK1-NEXT:    [[TE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TH:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 8
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 // CHECK1-NEXT:    [[N_CASTED7:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [3 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [3 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [3 x ptr], align 8
 // CHECK1-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[KERNEL_ARGS18:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 // CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
 // CHECK1-NEXT:    store i32 1000, ptr [[N]], align 4
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
 // CHECK1-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
 // CHECK1-NEXT:    store i32 128, ptr [[TH]], align 4
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
 // CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[I]], align 4
 // CHECK1-NEXT:    store i32 [[TMP3]], ptr [[I_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[I_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 [[TMP9]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK1-NEXT:    store i64 [[TMP4]], ptr [[TMP11]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK1-NEXT:    store i64 [[TMP4]], ptr [[TMP12]], align 8
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 // CHECK1-NEXT:    store ptr null, ptr [[TMP13]], align 8
 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 // CHECK1-NEXT:    store i64 [[TMP6]], ptr [[TMP14]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 // CHECK1-NEXT:    store i64 [[TMP6]], ptr [[TMP15]], align 8
 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
 // CHECK1-NEXT:    store ptr null, ptr [[TMP16]], align 8
 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP17]], align 8
 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP18]], align 8
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
 // CHECK1-NEXT:    store ptr null, ptr [[TMP19]], align 8
 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 // CHECK1-NEXT:    store i64 [[TMP8]], ptr [[TMP20]], align 8
 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 // CHECK1-NEXT:    store i64 [[TMP8]], ptr [[TMP21]], align 8
 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
 // CHECK1-NEXT:    store i64 [[TMP10]], ptr [[TMP23]], align 8
 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
 // CHECK1-NEXT:    store i64 [[TMP10]], ptr [[TMP24]], align 8
 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
 // CHECK1-NEXT:    store ptr null, ptr [[TMP25]], align 8
 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N]], align 4
 // CHECK1-NEXT:    store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR_3]], align 4
 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP31]], 0
 // CHECK1-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK1-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[DIV5]], 1
 // CHECK1-NEXT:    store i32 [[SUB6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP32]], 1
 // CHECK1-NEXT:    [[TMP33:%.*]] = zext i32 [[ADD]] to i64
 // CHECK1-NEXT:    [[TMP34:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP28]], 0
 // CHECK1-NEXT:    [[TMP35:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP29]], 0
 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK1-NEXT:    store i32 3, ptr [[TMP36]], align 4
 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 // CHECK1-NEXT:    store i32 5, ptr [[TMP37]], align 4
 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[TMP26]], ptr [[TMP38]], align 8
 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr [[TMP27]], ptr [[TMP39]], align 8
 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP40]], align 8
 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP41]], align 8
 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 // CHECK1-NEXT:    store ptr null, ptr [[TMP42]], align 8
 // CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 // CHECK1-NEXT:    store ptr null, ptr [[TMP43]], align 8
 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 // CHECK1-NEXT:    store i64 [[TMP33]], ptr [[TMP44]], align 8
 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 // CHECK1-NEXT:    store i64 0, ptr [[TMP45]], align 8
 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 // CHECK1-NEXT:    store [3 x i32] [[TMP34]], ptr [[TMP46]], align 4
 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 // CHECK1-NEXT:    store [3 x i32] [[TMP35]], ptr [[TMP47]], align 4
 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP48]], align 4
 // CHECK1-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4:[0-9]+]], i64 -1, i32 [[TMP28]], i32 [[TMP29]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
 // CHECK1-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50(i64 [[TMP4]], i64 [[TMP6]], ptr [[A]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR3:[0-9]+]]
 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 // CHECK1:       omp_offload.cont:
 // CHECK1-NEXT:    [[TMP51:%.*]] = load i32, ptr [[N]], align 4
 // CHECK1-NEXT:    store i32 [[TMP51]], ptr [[N_CASTED7]], align 4
 // CHECK1-NEXT:    [[TMP52:%.*]] = load i64, ptr [[N_CASTED7]], align 8
 // CHECK1-NEXT:    [[TMP53:%.*]] = load ptr, ptr [[G_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
 // CHECK1-NEXT:    store i64 [[TMP52]], ptr [[TMP54]], align 8
 // CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
 // CHECK1-NEXT:    store i64 [[TMP52]], ptr [[TMP55]], align 8
 // CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
 // CHECK1-NEXT:    store ptr null, ptr [[TMP56]], align 8
 // CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP57]], align 8
 // CHECK1-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP58]], align 8
 // CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
 // CHECK1-NEXT:    store ptr null, ptr [[TMP59]], align 8
 // CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[TMP53]], ptr [[TMP60]], align 8
 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[TMP53]], ptr [[TMP61]], align 8
 // CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 2
 // CHECK1-NEXT:    store ptr null, ptr [[TMP62]], align 8
 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP65:%.*]] = load i32, ptr [[N]], align 4
 // CHECK1-NEXT:    store i32 [[TMP65]], ptr [[DOTCAPTURE_EXPR_12]], align 4
 // CHECK1-NEXT:    [[TMP66:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_12]], align 4
 // CHECK1-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP66]], 0
 // CHECK1-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
 // CHECK1-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
 // CHECK1-NEXT:    store i32 [[SUB16]], ptr [[DOTCAPTURE_EXPR_13]], align 4
 // CHECK1-NEXT:    [[TMP67:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_13]], align 4
 // CHECK1-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP67]], 1
 // CHECK1-NEXT:    [[TMP68:%.*]] = zext i32 [[ADD17]] to i64
 // CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 0
 // CHECK1-NEXT:    store i32 3, ptr [[TMP69]], align 4
 // CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 1
 // CHECK1-NEXT:    store i32 3, ptr [[TMP70]], align 4
 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[TMP63]], ptr [[TMP71]], align 8
 // CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr [[TMP64]], ptr [[TMP72]], align 8
 // CHECK1-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 4
 // CHECK1-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP73]], align 8
 // CHECK1-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 5
 // CHECK1-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP74]], align 8
 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 6
 // CHECK1-NEXT:    store ptr null, ptr [[TMP75]], align 8
 // CHECK1-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 7
 // CHECK1-NEXT:    store ptr null, ptr [[TMP76]], align 8
 // CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 8
 // CHECK1-NEXT:    store i64 [[TMP68]], ptr [[TMP77]], align 8
 // CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 9
 // CHECK1-NEXT:    store i64 0, ptr [[TMP78]], align 8
 // CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 10
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP79]], align 4
 // CHECK1-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 11
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP80]], align 4
 // CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP81]], align 4
 // CHECK1-NEXT:    [[TMP82:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.region_id, ptr [[KERNEL_ARGS18]])
 // CHECK1-NEXT:    [[TMP83:%.*]] = icmp ne i32 [[TMP82]], 0
 // CHECK1-NEXT:    br i1 [[TMP83]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
 // CHECK1:       omp_offload.failed19:
 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56(i64 [[TMP52]], ptr [[A]], ptr [[TMP53]]) #[[ATTR3]]
 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
 // CHECK1:       omp_offload.cont20:
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i64 0, i64 0
 // CHECK1-NEXT:    [[TMP84:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    ret i32 [[TMP84]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50
 // CHECK1-SAME: (i64 noundef [[I:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[I_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[I_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB4]])
 // CHECK1-NEXT:    store i64 [[I]], ptr [[I_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB4]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[I_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[I_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED]], align 8
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined, i64 [[TMP5]], i64 [[TMP7]], ptr [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[I:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[I_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[I]], ptr [[I_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[I3]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 0
 // CHECK1-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i64 8) ]
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I4]], align 4
 // CHECK1-NEXT:    store i32 [[TMP18]], ptr [[I_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[I_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP20]], ptr [[N_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i64, ptr [[N_CASTED]], align 8
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]], ptr [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP25]])
 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
 // CHECK1-NEXT:    br i1 [[TMP27]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP28]], 0
 // CHECK1-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD9]], ptr [[I_ADDR]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[I:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[I_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I5:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I6:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[I]], ptr [[I_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[I3]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 0
 // CHECK1-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i64 8) ]
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTLINEAR_START]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTLINEAR_STEP]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[CONV4]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
 // CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2:[0-9]+]], i32 [[TMP10]])
 // CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP12]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK1-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP17]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
 // CHECK1-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I5]], align 4
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[I5]], align 4
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK1-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP24]])
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
 // CHECK1-NEXT:    br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB10:%.*]] = sub nsw i32 [[TMP27]], 0
 // CHECK1-NEXT:    [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
 // CHECK1-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
 // CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
 // CHECK1-NEXT:    store i32 [[ADD13]], ptr [[I_ADDR]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 // CHECK1-NEXT:    br i1 [[TMP29]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
 // CHECK1:       .omp.linear.pu:
 // CHECK1-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
 // CHECK1:       .omp.linear.pu.done:
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56
 // CHECK1-SAME: (i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[N_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 8
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined, i64 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12:![0-9]+]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[N_CASTED]], align 8, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[G_ADDR]], align 8, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], ptr [[TMP20]]), !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP24]])
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
 // CHECK1-NEXT:    br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP27]], 0
 // CHECK1-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD8]], ptr [[I3]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
 // CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16:![0-9]+]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 8, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i64 0
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 [[TMP18]], ptr [[ARRAYIDX7]], align 4, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
 // CHECK1-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP22]])
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
 // CHECK1-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP25]], 0
 // CHECK1-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
 // CHECK1-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
 // CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
 // CHECK1-NEXT:    store i32 [[ADD12]], ptr [[I4]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_Z24test_target_teams_atomicv
 // CHECK1-SAME: () #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[X:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[X]], align 4
 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK1-NEXT:    store ptr [[X]], ptr [[TMP0]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK1-NEXT:    store ptr [[X]], ptr [[TMP1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK1-NEXT:    store i32 3, ptr [[TMP5]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 // CHECK1-NEXT:    store i32 1, ptr [[TMP6]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 // CHECK1-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP9]], align 8
 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 // CHECK1-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP10]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 // CHECK1-NEXT:    store i64 1024, ptr [[TMP13]], align 8
 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 // CHECK1-NEXT:    store i64 0, ptr [[TMP14]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72(ptr [[X]]) #[[ATTR3]]
 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 // CHECK1:       omp_offload.cont:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 8
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined, ptr [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 1023, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1023
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1023, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP19:![0-9]+]]
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP19]]
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP19]]
 // CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP19]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]]), !llvm.access.group [[ACC_GRP19]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP19]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP19]]
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP19]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
 // CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 1023, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1023
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1023, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22:![0-9]+]]
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP22]]
 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]]
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP22]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = atomicrmw add ptr [[TMP0]], i32 1 monotonic, align 4, !llvm.access.group [[ACC_GRP22]]
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]]
 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP22]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP4]])
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
 // CHECK3-SAME: (ptr noundef [[G:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[N:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
 // CHECK3-NEXT:    [[TE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TH:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 // CHECK3-NEXT:    [[N_CASTED7:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [3 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [3 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [3 x ptr], align 4
 // CHECK3-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[KERNEL_ARGS18:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 // CHECK3-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
 // CHECK3-NEXT:    store i32 1000, ptr [[N]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
 // CHECK3-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
 // CHECK3-NEXT:    store i32 128, ptr [[TH]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[I]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], ptr [[I_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 [[TMP9]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[TMP11]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[TMP12]], align 4
 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 // CHECK3-NEXT:    store ptr null, ptr [[TMP13]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[TMP14]], align 4
 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[TMP15]], align 4
 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr null, ptr [[TMP16]], align 4
 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP17]], align 4
 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP18]], align 4
 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr null, ptr [[TMP19]], align 4
 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 // CHECK3-NEXT:    store i32 [[TMP8]], ptr [[TMP20]], align 4
 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 // CHECK3-NEXT:    store i32 [[TMP8]], ptr [[TMP21]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr null, ptr [[TMP22]], align 4
 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
 // CHECK3-NEXT:    store i32 [[TMP10]], ptr [[TMP23]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
 // CHECK3-NEXT:    store i32 [[TMP10]], ptr [[TMP24]], align 4
 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
 // CHECK3-NEXT:    store ptr null, ptr [[TMP25]], align 4
 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N]], align 4
 // CHECK3-NEXT:    store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR_3]], align 4
 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP31]], 0
 // CHECK3-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK3-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[DIV5]], 1
 // CHECK3-NEXT:    store i32 [[SUB6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP32]], 1
 // CHECK3-NEXT:    [[TMP33:%.*]] = zext i32 [[ADD]] to i64
 // CHECK3-NEXT:    [[TMP34:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP28]], 0
 // CHECK3-NEXT:    [[TMP35:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP29]], 0
 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 3, ptr [[TMP36]], align 4
 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 // CHECK3-NEXT:    store i32 5, ptr [[TMP37]], align 4
 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[TMP26]], ptr [[TMP38]], align 4
 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr [[TMP27]], ptr [[TMP39]], align 4
 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP40]], align 4
 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP41]], align 4
 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 // CHECK3-NEXT:    store ptr null, ptr [[TMP42]], align 4
 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 // CHECK3-NEXT:    store ptr null, ptr [[TMP43]], align 4
 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 // CHECK3-NEXT:    store i64 [[TMP33]], ptr [[TMP44]], align 8
 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 // CHECK3-NEXT:    store i64 0, ptr [[TMP45]], align 8
 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 // CHECK3-NEXT:    store [3 x i32] [[TMP34]], ptr [[TMP46]], align 4
 // CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 // CHECK3-NEXT:    store [3 x i32] [[TMP35]], ptr [[TMP47]], align 4
 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP48]], align 4
 // CHECK3-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4:[0-9]+]], i64 -1, i32 [[TMP28]], i32 [[TMP29]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
 // CHECK3-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50(i32 [[TMP4]], i32 [[TMP6]], ptr [[A]], i32 [[TMP8]], i32 [[TMP10]]) #[[ATTR3:[0-9]+]]
 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 // CHECK3:       omp_offload.cont:
 // CHECK3-NEXT:    [[TMP51:%.*]] = load i32, ptr [[N]], align 4
 // CHECK3-NEXT:    store i32 [[TMP51]], ptr [[N_CASTED7]], align 4
 // CHECK3-NEXT:    [[TMP52:%.*]] = load i32, ptr [[N_CASTED7]], align 4
 // CHECK3-NEXT:    [[TMP53:%.*]] = load ptr, ptr [[G_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 [[TMP52]], ptr [[TMP54]], align 4
 // CHECK3-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 [[TMP52]], ptr [[TMP55]], align 4
 // CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
 // CHECK3-NEXT:    store ptr null, ptr [[TMP56]], align 4
 // CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP57]], align 4
 // CHECK3-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP58]], align 4
 // CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr null, ptr [[TMP59]], align 4
 // CHECK3-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[TMP53]], ptr [[TMP60]], align 4
 // CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[TMP53]], ptr [[TMP61]], align 4
 // CHECK3-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr null, ptr [[TMP62]], align 4
 // CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP65:%.*]] = load i32, ptr [[N]], align 4
 // CHECK3-NEXT:    store i32 [[TMP65]], ptr [[DOTCAPTURE_EXPR_12]], align 4
 // CHECK3-NEXT:    [[TMP66:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_12]], align 4
 // CHECK3-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP66]], 0
 // CHECK3-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
 // CHECK3-NEXT:    [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
 // CHECK3-NEXT:    store i32 [[SUB16]], ptr [[DOTCAPTURE_EXPR_13]], align 4
 // CHECK3-NEXT:    [[TMP67:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_13]], align 4
 // CHECK3-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP67]], 1
 // CHECK3-NEXT:    [[TMP68:%.*]] = zext i32 [[ADD17]] to i64
 // CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 3, ptr [[TMP69]], align 4
 // CHECK3-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 1
 // CHECK3-NEXT:    store i32 3, ptr [[TMP70]], align 4
 // CHECK3-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[TMP63]], ptr [[TMP71]], align 4
 // CHECK3-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr [[TMP64]], ptr [[TMP72]], align 4
 // CHECK3-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 4
 // CHECK3-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP73]], align 4
 // CHECK3-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 5
 // CHECK3-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP74]], align 4
 // CHECK3-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 6
 // CHECK3-NEXT:    store ptr null, ptr [[TMP75]], align 4
 // CHECK3-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 7
 // CHECK3-NEXT:    store ptr null, ptr [[TMP76]], align 4
 // CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 8
 // CHECK3-NEXT:    store i64 [[TMP68]], ptr [[TMP77]], align 8
 // CHECK3-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 9
 // CHECK3-NEXT:    store i64 0, ptr [[TMP78]], align 8
 // CHECK3-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 10
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP79]], align 4
 // CHECK3-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 11
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP80]], align 4
 // CHECK3-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS18]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP81]], align 4
 // CHECK3-NEXT:    [[TMP82:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.region_id, ptr [[KERNEL_ARGS18]])
 // CHECK3-NEXT:    [[TMP83:%.*]] = icmp ne i32 [[TMP82]], 0
 // CHECK3-NEXT:    br i1 [[TMP83]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
 // CHECK3:       omp_offload.failed19:
 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56(i32 [[TMP52]], ptr [[A]], ptr [[TMP53]]) #[[ATTR3]]
 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
 // CHECK3:       omp_offload.cont20:
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP84:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    ret i32 [[TMP84]]
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50
 // CHECK3-SAME: (i32 noundef [[I:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[I_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB4]])
 // CHECK3-NEXT:    store i32 [[I]], ptr [[I_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB4]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[I_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[N_CASTED]], align 4
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined, i32 [[TMP5]], i32 [[TMP7]], ptr [[TMP1]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[I:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[I_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[I]], ptr [[I_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[I3]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK3:       omp.precond.then:
 // CHECK3-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i32 8) ]
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I4]], align 4
 // CHECK3-NEXT:    store i32 [[TMP16]], ptr [[I_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N_CASTED]], align 4
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined.omp_outlined, i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], i32 [[TMP19]], ptr [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
 // CHECK3-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP26]], 0
 // CHECK3-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD9]], ptr [[I_ADDR]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       omp.precond.end:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[I:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I5:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[I]], ptr [[I_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[I3]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK3:       omp.precond.then:
 // CHECK3-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i32 8) ]
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTLINEAR_START]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTLINEAR_STEP]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
 // CHECK3-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2:[0-9]+]], i32 [[TMP10]])
 // CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP12]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK3-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP17]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
 // CHECK3-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[I4]], align 4
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 [[TMP21]]
 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP24]])
 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
 // CHECK3-NEXT:    br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP27]], 0
 // CHECK3-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
 // CHECK3-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
 // CHECK3-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
 // CHECK3-NEXT:    store i32 [[ADD12]], ptr [[I_ADDR]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 // CHECK3-NEXT:    br i1 [[TMP29]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
 // CHECK3:       .omp.linear.pu:
 // CHECK3-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
 // CHECK3:       .omp.linear.pu.done:
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       omp.precond.end:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56
 // CHECK3-SAME: (i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 4
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined, i32 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK3:       omp.precond.then:
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13:![0-9]+]]
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    store i32 [[TMP16]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[G_ADDR]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined.omp_outlined, i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], ptr [[TMP0]], ptr [[TMP18]]), !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
 // CHECK3-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP25]], 0
 // CHECK3-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[I3]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       omp.precond.end:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK3:       omp.precond.then:
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17:![0-9]+]]
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I3]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 [[TMP19]]
 // CHECK3-NEXT:    store i32 [[TMP18]], ptr [[ARRAYIDX6]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
 // CHECK3-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP22]])
 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
 // CHECK3-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
 // CHECK3-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
 // CHECK3-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
 // CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
 // CHECK3-NEXT:    store i32 [[ADD11]], ptr [[I3]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       omp.precond.end:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z24test_target_teams_atomicv
 // CHECK3-SAME: () #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[X:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 // CHECK3-NEXT:    store i32 0, ptr [[X]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK3-NEXT:    store ptr [[X]], ptr [[TMP0]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK3-NEXT:    store ptr [[X]], ptr [[TMP1]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 // CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 3, ptr [[TMP5]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 // CHECK3-NEXT:    store i32 1, ptr [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 // CHECK3-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP9]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 // CHECK3-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP10]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 // CHECK3-NEXT:    store i64 1024, ptr [[TMP13]], align 8
 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 // CHECK3-NEXT:    store i64 0, ptr [[TMP14]], align 8
 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72(ptr [[X]]) #[[ATTR3]]
 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 // CHECK3:       omp_offload.cont:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined, ptr [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 1023, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1023
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1023, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20:![0-9]+]]
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]]), !llvm.access.group [[ACC_GRP20]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
 // CHECK3-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 1023, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1023
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1023, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23:![0-9]+]]
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK3-NEXT:    [[TMP11:%.*]] = atomicrmw add ptr [[TMP0]], i32 1 monotonic, align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK3-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP4]])
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
 // CHECK5-SAME: (ptr noundef [[G:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[N:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
 // CHECK5-NEXT:    [[TE:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[TH:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I6:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I7:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I8:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[_TMP15:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_16:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_17:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I23:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_IV26:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I27:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
 // CHECK5-NEXT:    store i32 1000, ptr [[N]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 // CHECK5-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
 // CHECK5-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
 // CHECK5-NEXT:    store i32 128, ptr [[TH]], align 4
 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
 // CHECK5-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
 // CHECK5-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
 // CHECK5-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
 // CHECK5-NEXT:    [[DIV4:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK5-NEXT:    [[SUB5:%.*]] = sub nsw i32 [[DIV4]], 1
 // CHECK5-NEXT:    store i32 [[SUB5]], ptr [[DOTCAPTURE_EXPR_3]], align 4
 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
 // CHECK5-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    store i32 0, ptr [[I6]], align 4
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK5-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
 // CHECK5-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
 // CHECK5:       simd.if.then:
 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i64 0, i64 0
 // CHECK5-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i64 8) ]
 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[I]], align 4
 // CHECK5-NEXT:    store i32 [[TMP8]], ptr [[DOTLINEAR_START]], align 4
 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[N]], align 4
 // CHECK5-NEXT:    store i32 [[TMP9]], ptr [[DOTLINEAR_STEP]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK5:       omp.inner.for.cond:
 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK5-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I7]], align 4
 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I7]], align 4
 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK5-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK5:       omp.body.continue:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK5:       omp.inner.for.inc:
 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP14]], 1
 // CHECK5-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
 // CHECK5:       omp.inner.for.end:
 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK5-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP15]], 0
 // CHECK5-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
 // CHECK5-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[DIV12]], 1
 // CHECK5-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
 // CHECK5-NEXT:    store i32 [[ADD14]], ptr [[I]], align 4
 // CHECK5-NEXT:    br label [[SIMD_IF_END]]
 // CHECK5:       simd.if.end:
 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N]], align 4
 // CHECK5-NEXT:    store i32 [[TMP16]], ptr [[DOTCAPTURE_EXPR_16]], align 4
 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_16]], align 4
 // CHECK5-NEXT:    [[SUB18:%.*]] = sub nsw i32 [[TMP17]], 0
 // CHECK5-NEXT:    [[DIV19:%.*]] = sdiv i32 [[SUB18]], 1
 // CHECK5-NEXT:    [[SUB20:%.*]] = sub nsw i32 [[DIV19]], 1
 // CHECK5-NEXT:    store i32 [[SUB20]], ptr [[DOTCAPTURE_EXPR_17]], align 4
 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB21]], align 4
 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_17]], align 4
 // CHECK5-NEXT:    store i32 [[TMP18]], ptr [[DOTOMP_UB22]], align 4
 // CHECK5-NEXT:    store i32 0, ptr [[I23]], align 4
 // CHECK5-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_16]], align 4
 // CHECK5-NEXT:    [[CMP24:%.*]] = icmp slt i32 0, [[TMP19]]
 // CHECK5-NEXT:    br i1 [[CMP24]], label [[SIMD_IF_THEN25:%.*]], label [[SIMD_IF_END44:%.*]]
 // CHECK5:       simd.if.then25:
 // CHECK5-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_LB21]], align 4
 // CHECK5-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV26]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND28:%.*]]
 // CHECK5:       omp.inner.for.cond28:
 // CHECK5-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV26]], align 4, !llvm.access.group [[ACC_GRP5:![0-9]+]]
 // CHECK5-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_UB22]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    [[CMP29:%.*]] = icmp sle i32 [[TMP21]], [[TMP22]]
 // CHECK5-NEXT:    br i1 [[CMP29]], label [[OMP_INNER_FOR_BODY30:%.*]], label [[OMP_INNER_FOR_END39:%.*]]
 // CHECK5:       omp.inner.for.body30:
 // CHECK5-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV26]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    [[MUL31:%.*]] = mul nsw i32 [[TMP23]], 1
 // CHECK5-NEXT:    [[ADD32:%.*]] = add nsw i32 0, [[MUL31]]
 // CHECK5-NEXT:    store i32 [[ADD32]], ptr [[I27]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[G_ADDR]], align 8, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    [[ARRAYIDX33:%.*]] = getelementptr inbounds i32, ptr [[TMP24]], i64 0
 // CHECK5-NEXT:    [[TMP25:%.*]] = load i32, ptr [[ARRAYIDX33]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    [[TMP26:%.*]] = load i32, ptr [[I27]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    [[IDXPROM34:%.*]] = sext i32 [[TMP26]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX35:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i64 0, i64 [[IDXPROM34]]
 // CHECK5-NEXT:    store i32 [[TMP25]], ptr [[ARRAYIDX35]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE36:%.*]]
 // CHECK5:       omp.body.continue36:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
 // CHECK5:       omp.inner.for.inc37:
 // CHECK5-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_IV26]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP27]], 1
 // CHECK5-NEXT:    store i32 [[ADD38]], ptr [[DOTOMP_IV26]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND28]], !llvm.loop [[LOOP6:![0-9]+]]
 // CHECK5:       omp.inner.for.end39:
 // CHECK5-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_16]], align 4
 // CHECK5-NEXT:    [[SUB40:%.*]] = sub nsw i32 [[TMP28]], 0
 // CHECK5-NEXT:    [[DIV41:%.*]] = sdiv i32 [[SUB40]], 1
 // CHECK5-NEXT:    [[MUL42:%.*]] = mul nsw i32 [[DIV41]], 1
 // CHECK5-NEXT:    [[ADD43:%.*]] = add nsw i32 0, [[MUL42]]
 // CHECK5-NEXT:    store i32 [[ADD43]], ptr [[I27]], align 4
 // CHECK5-NEXT:    br label [[SIMD_IF_END44]]
 // CHECK5:       simd.if.end44:
 // CHECK5-NEXT:    [[ARRAYIDX45:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i64 0, i64 0
 // CHECK5-NEXT:    [[TMP29:%.*]] = load i32, ptr [[ARRAYIDX45]], align 4
 // CHECK5-NEXT:    ret i32 [[TMP29]]
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_Z24test_target_teams_atomicv
 // CHECK5-SAME: () #[[ATTR0]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[X:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    store i32 0, ptr [[X]], align 4
 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 1023, ptr [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK5:       omp.inner.for.cond:
 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[X]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
 // CHECK5-NEXT:    store i32 [[ADD1]], ptr [[X]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK5:       omp.body.continue:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK5:       omp.inner.for.inc:
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
 // CHECK5-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
 // CHECK5:       omp.inner.for.end:
 // CHECK5-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_Z16target_teams_funPi
 // CHECK7-SAME: (ptr noundef [[G:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    [[N:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[A:%.*]] = alloca [1000 x i32], align 4
 // CHECK7-NEXT:    [[TE:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[TH:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I6:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I7:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I8:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[_TMP15:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_16:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_17:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I23:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_IV26:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I27:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
 // CHECK7-NEXT:    store i32 1000, ptr [[N]], align 4
 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 // CHECK7-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
 // CHECK7-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
 // CHECK7-NEXT:    store i32 128, ptr [[TH]], align 4
 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
 // CHECK7-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
 // CHECK7-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
 // CHECK7-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK7-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
 // CHECK7-NEXT:    [[DIV4:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK7-NEXT:    [[SUB5:%.*]] = sub nsw i32 [[DIV4]], 1
 // CHECK7-NEXT:    store i32 [[SUB5]], ptr [[DOTCAPTURE_EXPR_3]], align 4
 // CHECK7-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
 // CHECK7-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    store i32 0, ptr [[I6]], align 4
 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK7-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
 // CHECK7-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
 // CHECK7:       simd.if.then:
 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i32 0, i32 0
 // CHECK7-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i32 8) ]
 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, ptr [[I]], align 4
 // CHECK7-NEXT:    store i32 [[TMP8]], ptr [[DOTLINEAR_START]], align 4
 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, ptr [[N]], align 4
 // CHECK7-NEXT:    store i32 [[TMP9]], ptr [[DOTLINEAR_STEP]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK7:       omp.inner.for.cond:
 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK7-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK7:       omp.inner.for.body:
 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK7-NEXT:    store i32 [[ADD]], ptr [[I7]], align 4
 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I7]], align 4
 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i32 0, i32 [[TMP13]]
 // CHECK7-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK7:       omp.body.continue:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK7:       omp.inner.for.inc:
 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP14]], 1
 // CHECK7-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
 // CHECK7:       omp.inner.for.end:
 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK7-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP15]], 0
 // CHECK7-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
 // CHECK7-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[DIV12]], 1
 // CHECK7-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
 // CHECK7-NEXT:    store i32 [[ADD14]], ptr [[I]], align 4
 // CHECK7-NEXT:    br label [[SIMD_IF_END]]
 // CHECK7:       simd.if.end:
 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N]], align 4
 // CHECK7-NEXT:    store i32 [[TMP16]], ptr [[DOTCAPTURE_EXPR_16]], align 4
 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_16]], align 4
 // CHECK7-NEXT:    [[SUB18:%.*]] = sub nsw i32 [[TMP17]], 0
 // CHECK7-NEXT:    [[DIV19:%.*]] = sdiv i32 [[SUB18]], 1
 // CHECK7-NEXT:    [[SUB20:%.*]] = sub nsw i32 [[DIV19]], 1
 // CHECK7-NEXT:    store i32 [[SUB20]], ptr [[DOTCAPTURE_EXPR_17]], align 4
 // CHECK7-NEXT:    store i32 0, ptr [[DOTOMP_LB21]], align 4
 // CHECK7-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_17]], align 4
 // CHECK7-NEXT:    store i32 [[TMP18]], ptr [[DOTOMP_UB22]], align 4
 // CHECK7-NEXT:    store i32 0, ptr [[I23]], align 4
 // CHECK7-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_16]], align 4
 // CHECK7-NEXT:    [[CMP24:%.*]] = icmp slt i32 0, [[TMP19]]
 // CHECK7-NEXT:    br i1 [[CMP24]], label [[SIMD_IF_THEN25:%.*]], label [[SIMD_IF_END43:%.*]]
 // CHECK7:       simd.if.then25:
 // CHECK7-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_LB21]], align 4
 // CHECK7-NEXT:    store i32 [[TMP20]], ptr [[DOTOMP_IV26]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND28:%.*]]
 // CHECK7:       omp.inner.for.cond28:
 // CHECK7-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV26]], align 4, !llvm.access.group [[ACC_GRP6:![0-9]+]]
 // CHECK7-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_UB22]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    [[CMP29:%.*]] = icmp sle i32 [[TMP21]], [[TMP22]]
 // CHECK7-NEXT:    br i1 [[CMP29]], label [[OMP_INNER_FOR_BODY30:%.*]], label [[OMP_INNER_FOR_END38:%.*]]
 // CHECK7:       omp.inner.for.body30:
 // CHECK7-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV26]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    [[MUL31:%.*]] = mul nsw i32 [[TMP23]], 1
 // CHECK7-NEXT:    [[ADD32:%.*]] = add nsw i32 0, [[MUL31]]
 // CHECK7-NEXT:    store i32 [[ADD32]], ptr [[I27]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[G_ADDR]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    [[ARRAYIDX33:%.*]] = getelementptr inbounds i32, ptr [[TMP24]], i32 0
 // CHECK7-NEXT:    [[TMP25:%.*]] = load i32, ptr [[ARRAYIDX33]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    [[TMP26:%.*]] = load i32, ptr [[I27]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i32 0, i32 [[TMP26]]
 // CHECK7-NEXT:    store i32 [[TMP25]], ptr [[ARRAYIDX34]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE35:%.*]]
 // CHECK7:       omp.body.continue35:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC36:%.*]]
 // CHECK7:       omp.inner.for.inc36:
 // CHECK7-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_IV26]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    [[ADD37:%.*]] = add nsw i32 [[TMP27]], 1
 // CHECK7-NEXT:    store i32 [[ADD37]], ptr [[DOTOMP_IV26]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND28]], !llvm.loop [[LOOP7:![0-9]+]]
 // CHECK7:       omp.inner.for.end38:
 // CHECK7-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_16]], align 4
 // CHECK7-NEXT:    [[SUB39:%.*]] = sub nsw i32 [[TMP28]], 0
 // CHECK7-NEXT:    [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
 // CHECK7-NEXT:    [[MUL41:%.*]] = mul nsw i32 [[DIV40]], 1
 // CHECK7-NEXT:    [[ADD42:%.*]] = add nsw i32 0, [[MUL41]]
 // CHECK7-NEXT:    store i32 [[ADD42]], ptr [[I27]], align 4
 // CHECK7-NEXT:    br label [[SIMD_IF_END43]]
 // CHECK7:       simd.if.end43:
 // CHECK7-NEXT:    [[ARRAYIDX44:%.*]] = getelementptr inbounds [1000 x i32], ptr [[A]], i32 0, i32 0
 // CHECK7-NEXT:    [[TMP29:%.*]] = load i32, ptr [[ARRAYIDX44]], align 4
 // CHECK7-NEXT:    ret i32 [[TMP29]]
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_Z24test_target_teams_atomicv
 // CHECK7-SAME: () #[[ATTR0]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[X:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    store i32 0, ptr [[X]], align 4
 // CHECK7-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    store i32 1023, ptr [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK7:       omp.inner.for.cond:
 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10:![0-9]+]]
 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK7:       omp.inner.for.body:
 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK7-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, ptr [[X]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
 // CHECK7-NEXT:    store i32 [[ADD1]], ptr [[X]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK7:       omp.body.continue:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK7:       omp.inner.for.inc:
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
 // CHECK7-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
 // CHECK7:       omp.inner.for.end:
 // CHECK7-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50
 // CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[I:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[I_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[I_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB4:[0-9]+]])
 // CHECK9-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[I]], ptr [[I_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB4]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[I_CASTED]], align 4
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, ptr [[I_CASTED]], align 8
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED]], align 4
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED]], align 8
 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined, i64 [[TMP5]], i64 [[TMP7]], ptr [[TMP1]])
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined
 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[I:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[I_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[I]], ptr [[I_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[I3]], align 4
 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK9:       omp.precond.then:
 // CHECK9-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 0
 // CHECK9-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i64 8) ]
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I4]], align 4
 // CHECK9-NEXT:    store i32 [[TMP18]], ptr [[I_CASTED]], align 4
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[I_CASTED]], align 8
 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP20]], ptr [[N_CASTED]], align 4
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i64, ptr [[N_CASTED]], align 8
 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]], ptr [[TMP0]])
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
 // CHECK9-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP25]])
 // CHECK9-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
 // CHECK9-NEXT:    br i1 [[TMP27]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP28]], 0
 // CHECK9-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD9]], ptr [[I_ADDR]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined.omp_outlined
 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[I:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[I_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I5:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I6:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[I]], ptr [[I_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[I3]], align 4
 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK9:       omp.precond.then:
 // CHECK9-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 0
 // CHECK9-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i64 8) ]
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTLINEAR_START]], align 4
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTLINEAR_STEP]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK9-NEXT:    [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
 // CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[CONV4]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
 // CHECK9-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2:[0-9]+]], i32 [[TMP10]])
 // CHECK9-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP12]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK9-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP17]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
 // CHECK9-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I5]], align 4
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[I5]], align 4
 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK9-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
 // CHECK9-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP24]])
 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
 // CHECK9-NEXT:    br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB10:%.*]] = sub nsw i32 [[TMP27]], 0
 // CHECK9-NEXT:    [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
 // CHECK9-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
 // CHECK9-NEXT:    [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
 // CHECK9-NEXT:    store i32 [[ADD13]], ptr [[I_ADDR]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 // CHECK9-NEXT:    br i1 [[TMP29]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
 // CHECK9:       .omp.linear.pu:
 // CHECK9-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
 // CHECK9:       .omp.linear.pu.done:
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56
 // CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[N_CASTED]], align 8
 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 8
 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined, i64 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined
 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK9:       omp.precond.then:
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK9-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13:![0-9]+]]
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK9-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[N_CASTED]], align 8, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[G_ADDR]], align 8, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined.omp_outlined, i64 [[TMP15]], i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], ptr [[TMP20]]), !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
 // CHECK9-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP24]])
 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
 // CHECK9-NEXT:    br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP27]], 0
 // CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD8]], ptr [[I3]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined.omp_outlined
 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK9:       omp.precond.then:
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK9-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
 // CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
 // CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17:![0-9]+]]
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 8, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i64 0
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK9-NEXT:    store i32 [[TMP18]], ptr [[ARRAYIDX7]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
 // CHECK9-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
 // CHECK9-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP22]])
 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
 // CHECK9-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK9-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP25]], 0
 // CHECK9-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
 // CHECK9-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
 // CHECK9-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
 // CHECK9-NEXT:    store i32 [[ADD12]], ptr [[I4]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72
 // CHECK9-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 8
 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined, ptr [[TMP0]])
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined
 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 8
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    store i32 1023, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1023
 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1023, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20:![0-9]+]]
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]]), !llvm.access.group [[ACC_GRP20]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP20]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined.omp_outlined
 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 8
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 1023, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
 // CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1023
 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1023, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23:![0-9]+]]
 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK9-NEXT:    [[TMP11:%.*]] = atomicrmw add ptr [[TMP0]], i32 1 monotonic, align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP23]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP4]])
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50
 // CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[I:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[I_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I_CASTED:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB4:[0-9]+]])
 // CHECK11-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[I]], ptr [[I_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB4]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[I_CASTED]], align 4
 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I_CASTED]], align 4
 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED]], align 4
 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[N_CASTED]], align 4
 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined, i32 [[TMP5]], i32 [[TMP7]], ptr [[TMP1]])
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined
 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[I:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[I_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I_CASTED:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[I]], ptr [[I_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[I3]], align 4
 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK11:       omp.precond.then:
 // CHECK11-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 0
 // CHECK11-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i32 8) ]
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:
 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK11-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I4]], align 4
 // CHECK11-NEXT:    store i32 [[TMP16]], ptr [[I_CASTED]], align 4
 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I_CASTED]], align 4
 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP18]], ptr [[N_CASTED]], align 4
 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N_CASTED]], align 4
 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined.omp_outlined, i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], i32 [[TMP19]], ptr [[TMP0]])
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
 // CHECK11-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
 // CHECK11-NEXT:    br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK11:       .omp.final.then:
 // CHECK11-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP26]], 0
 // CHECK11-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV8]], 1
 // CHECK11-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK11-NEXT:    store i32 [[ADD9]], ptr [[I_ADDR]], align 4
 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK11:       .omp.final.done:
 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK11:       omp.precond.end:
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l50.omp_outlined.omp_outlined
 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[I:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I5:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[I]], ptr [[I_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[I3]], align 4
 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK11:       omp.precond.then:
 // CHECK11-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 0
 // CHECK11-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[ARRAYDECAY]], i32 8) ]
 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTLINEAR_START]], align 4
 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTLINEAR_STEP]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
 // CHECK11-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2:[0-9]+]], i32 [[TMP10]])
 // CHECK11-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP12]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP13]], [[TMP14]]
 // CHECK11-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:
 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP17]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
 // CHECK11-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[I4]], align 4
 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 [[TMP21]]
 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK11:       omp.body.continue:
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
 // CHECK11-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
 // CHECK11-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP24]])
 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
 // CHECK11-NEXT:    br i1 [[TMP26]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK11:       .omp.final.then:
 // CHECK11-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i32 [[TMP27]], 0
 // CHECK11-NEXT:    [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
 // CHECK11-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
 // CHECK11-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
 // CHECK11-NEXT:    store i32 [[ADD12]], ptr [[I_ADDR]], align 4
 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK11:       .omp.final.done:
 // CHECK11-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 // CHECK11-NEXT:    br i1 [[TMP29]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
 // CHECK11:       .omp.linear.pu:
 // CHECK11-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
 // CHECK11:       .omp.linear.pu.done:
 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK11:       omp.precond.end:
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56
 // CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[N_CASTED]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_CASTED]], align 4
 // CHECK11-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[G_ADDR]], align 4
 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined, i32 [[TMP2]], ptr [[TMP0]], ptr [[TMP3]])
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined
 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK11:       omp.precond.then:
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
 // CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:
 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP9]], [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14:![0-9]+]]
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    store i32 [[TMP16]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[G_ADDR]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined.omp_outlined, i32 [[TMP14]], i32 [[TMP15]], i32 [[TMP17]], ptr [[TMP0]], ptr [[TMP18]]), !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
 // CHECK11-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
 // CHECK11-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK11:       .omp.final.then:
 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP25]], 0
 // CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK11-NEXT:    store i32 [[ADD8]], ptr [[I3]], align 4
 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK11:       .omp.final.done:
 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK11:       omp.precond.end:
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16target_teams_funPi_l56.omp_outlined.omp_outlined
 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4000) [[A:%.*]], ptr noundef [[G:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK11:       omp.precond.then:
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
 // CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:
 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]]
 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[G_ADDR]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I3]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [1000 x i32], ptr [[TMP0]], i32 0, i32 [[TMP19]]
 // CHECK11-NEXT:    store i32 [[TMP18]], ptr [[ARRAYIDX6]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK11:       omp.body.continue:
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
 // CHECK11-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
 // CHECK11-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP22]])
 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
 // CHECK11-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK11:       .omp.final.then:
 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 // CHECK11-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
 // CHECK11-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
 // CHECK11-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
 // CHECK11-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
 // CHECK11-NEXT:    store i32 [[ADD11]], ptr [[I3]], align 4
 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK11:       .omp.final.done:
 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK11:       omp.precond.end:
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72
 // CHECK11-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4
 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined, ptr [[TMP0]])
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined
 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK11-NEXT:    store i32 1023, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1023
 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:
 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1023, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21:![0-9]+]]
 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP21]]
 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP21]]
 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP21]]
 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]]), !llvm.access.group [[ACC_GRP21]]
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP21]]
 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
 // CHECK11-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK11:       .omp.final.then:
 // CHECK11-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK11:       .omp.final.done:
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z24test_target_teams_atomicv_l72.omp_outlined.omp_outlined
 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[X_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    store ptr [[X]], ptr [[X_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    store i32 1023, ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1023
 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK11:       cond.true:
 // CHECK11-NEXT:    br label [[COND_END:%.*]]
 // CHECK11:       cond.false:
 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    br label [[COND_END]]
 // CHECK11:       cond.end:
 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 1023, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24:![0-9]+]]
 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP24]]
 // CHECK11-NEXT:    [[TMP11:%.*]] = atomicrmw add ptr [[TMP0]], i32 1 monotonic, align 4, !llvm.access.group [[ACC_GRP24]]
 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK11:       omp.body.continue:
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK11-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP4]])
 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK11-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK11:       .omp.final.then:
 // CHECK11-NEXT:    store i32 1024, ptr [[I]], align 4
 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK11:       .omp.final.done:
 // CHECK11-NEXT:    ret void
 //